Early Screening of Autism Spectrum Disorder: Recommendations for Practice and Research - Family Outreach
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
SUPPLEMENT ARTICLE
Early Screening of Autism Spectrum Disorder:
Recommendations for Practice and Research
AUTHORS: Lonnie Zwaigenbaum, MD,a Margaret L.
Bauman, MD,b Deborah Fein, PhD,c Karen Pierce, PhD,d abstract
Timothy Buie, MD,e Patricia A. Davis, MD,f Craig
This article reviews current evidence for autism spectrum disorder
Newschaffer, PhD,g Diana L. Robins, PhD,g Amy Wetherby,
PhD,h Roula Choueiri, MD,i Connie Kasari, PhD,j Wendy L. (ASD) screening based on peer-reviewed articles published to Decem-
Stone, PhD,k Nurit Yirmiya, PhD,l Annette Estes, PhD,m ber 2013. Screening provides a standardized process to ensure that
Robin L. Hansen, MD,n James C. McPartland, PhD,o Marvin children are systematically monitored for early signs of ASD to pro-
R. Natowicz, MD, PhD,p Alice Carter, PhD,q Doreen mote earlier diagnosis. The current review indicates that screening
Granpeesheh, PhD, BCBA-D,r Zoe Mailloux, OTD, OTR/L, in children aged 18 to 24 months can assist in early detection, consis-
FAOTA,s Susanne Smith Roley, OTD, OTR/L, FAOTA,t and
Sheldon Wagner, PhDu
tent with current American Academy of Pediatrics’ recommendations.
aDepartment of Pediatrics, University of Alberta, Edmonton,
We identify ASD-specific and broadband screening tools that have
Alberta, Canada; bDepartment of Anatomy and Neurobiology, been ev-aluated in large community samples which show particular
Boston University School of Medicine, Boston, Massachusetts; promise in terms of accurate classification and clinical utility. We also
cDepartment of Psychology, University of Connecticut, Storrs,
suggest strategies to help overcome challenges to implementing ASD
Connecticut; dDepartment of Neurosciences, University of
California San Diego, La Jolla, California; eHarvard Medical screening in community practice, as well as priorities for future re-
School and Massachusetts General Hospital for Children, Boston, search. Pediatrics 2015;136:S41–S59
Massachusetts; fIntegrated Center for Child Development,
Newton, Massachusetts; gA.J. Drexel Autism Institute, Drexel
University, Philadelphia, Pennsylvania; hDepartment of Clinical
Sciences, Florida State University College of Medicine,
Tallahassee, Florida; iDivision of Developmental and Behavioral
Pediatrics, University of Massachusetts Memorial Children’s
Medical Center, Worcester, Massachusetts; jGraduate School of
Education & Information Studies, University of California Los
Angeles, Los Angeles, California; kDepartments of Psychology and
mSpeech and Hearing Sciences, University of Washington, Seattle,
Washington; lDepartment of Psychology, Hebrew University of
Jerusalem Mount Scopus, Jerusalem, Israel; nDepartment of
Pediatrics, University of California Davis MIND Institute,
Sacramento, California; oYale Child Study Center, New Haven,
Connecticut; pGenomic Medicine Institute, Cleveland Clinic,
Cleveland, Ohio; qDepartment of Psychology, University of
Massachusetts, Boston, Massachusetts; rCenter for Autism and
Related Disorders, Tarzana, California; sDepartment of
Occupational Therapy, Thomas Jefferson University, Philadelphia,
Pennsylvania; tUSC Mrs T.H. Chan Division of Occupational Science
and Occupational Therapy, Los Angeles, California; and
uBehavioral Development & Educational Services, New Bedford,
Massachusetts
(Continued on last page)
PEDIATRICS Volume 136, Supplement 1, October 2015 S41
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018Although there have been consider- practice has been modest.10,11 Although yet screens positive is considered to be
able advances in characterizing early potential facilitatorsand barriers to a false-positive. It has been suggested
behavioral markers predictive of au- ASD screening have been researched that to even receive consideration for
tism spectrum disorders (ASDs), as and debated,11–13 screening rates in many population screening applications, the
summarized in this special issue to regions of the United States remain low. sensitivity and specificity of a screen-
Pediatrics,1 translation into clinical Community-based interventions aimed ing tool should exceed 0.70.19 However,
practice requires that the process of at implementing or increasing utiliza- the relative “cost” associated with
monitoring for such early risk mark- tion of ASD screening have emphasized false-positive and false-negative find-
ers be operationalized to facilitate training primary care physicians and ings, as well as the prevalence of the
broad implementation. To that end, their front-line staff, providing ongoing condition being screened, must also be
universal screening for ASD has been technical assistance (eg, scoring, data taken into consideration. The positive
recommended by the American management support), and clear re- predictive value (PPV) for ASD of
Academy of Pediatrics (AAP) to ensure ferral pathways for specialized assess- a screening test is defined as the pro-
consistent practice and optimal de- ments.9,11,14–17 However, ongoing debate portion of children screening positive
tection of young children with early regarding whether there is sufficient who receive an ASD diagnosis divided
signs of ASD across a range of clinical evidence in support of ASD screening to by the total number of screen-positive
and community contexts.2 The AAP has warrant widespread practice change8,18 cases. The negative predictive value
recommended that all children be may undermine the degree to which (NPV) is the proportion of screen-
screened with an ASD-specific in- community pediatricians are adopting negative children not receiving an
strument during well-child visits at the AAP policy. ASD diagnosis. PPV and NPV are influ-
ages 18 and 24 months in conjunction enced by the baseline prevalence of
Thus, an updated literature review and
with ongoing developmental surveil- ASD in the population being screened
best practice recommendations re-
lance and broadband developmental as well as the sensitivity and specificity
garding ASD screening are warranted,
screening. The rationale for this rec- of the screening tool. Although sensi-
as well as further considerations of how
ommendation was based on the tivity and specificity are intrinsic
to address potential barriers to uptake
measures of test performance, PPV
presence of ASD symptoms by age 18 of screening into clinical practice. To
and NPV arguably have more inherent
months, promising data on early ASD- that end, an international multidisci-
meaning for individual family-level and
screening tools, and the availability of plinary panel of clinical practitioners
system-level evaluations of screening.
effective intervention strategies tar- and researchers with expertise in ASD
geting this age group.3,4 Recent ran- and developmental disabilities was It is also important to distinguish level 1
domized controlled trials have added convened in Marina del Rey, California from level 2 screening. Level 1 screen-
new evidence that for many children in October 2010. The panel reached ing applies to all children regardless of
aged ,3 years, early intervention can consensus on “How can we optimize risk status (ie, “universal” screening).
improve outcomes, including core developmental course and outcomes In contrast, level 2 screening is tar-
deficits of ASD (ie, social attention), IQ, through ASD screening programs for geted at children already identified as
language, and symptom severity,5,6 children aged #24 months?” being at increased risk (eg, due to
thus increasing the potential benefits a positive family history, concerns
For further context, we briefly define raised by parents or clinicians, identi-
of early diagnosis facilitated by early terms used to describe the classifica-
screening. fication by a level 1 screener).
tion accuracy of specific screening
Some scientists and practitioners have measures. “Sensitivity” refers to the
questioned whether the evidence rela- proportion of children with ASD who METHODS
tive to general developmental surveil- are correctly identified as “high risk” The working group co-chairs and panel
lance warrants ASD screening,7,8 and according to results of screening; co-chairs conducted a PubMed search
others have argued that research needs a child with ASD who is not identified by to identify relevant articles on screen-
to move beyond risk classification and the screen is considered to be a false- ing for ASD in children aged #24
evaluate longer term outcomes of ASD negative. Specificity refers to the pro- months. Members of the working
screening (eg, impact on age of di- portion of children who do not have group reviewed the articles. We
agnosis, related gains attributable to ASD who are correctly classified using assessed whether tools were being
earlier enrollment in intervention).9 The the screening tool as not having risk evaluated in the population in which
uptake of ASD screening into pediatric for ASD; a child who does not have ASD they were being considered for use and
S42 ZWAIGENBAUM et al
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018SUPPLEMENT ARTICLE
whether these tools met the minimum During the conference, the working SUMMARY STATEMENTS
criteria for specificity and sensitivity to group offered draft recommendations
Statement 1: Evidence supports the
support implementation in the general for discussion, modification, and rati- usefulness of ASD-specific
community. Panel recommendations fication by all attendees. Electronic screening at 18 and 24 months.
were based on this evaluative frame- voting was used to express opinions ASD screening before 24 months
work. and guide consensus building. A may be associated with higher
The working group summarized pub- modified nominal group technique was false-positive rates than screening
lished research on screening tools de- used to review the recommendations, at ‡24 months but may still be
veloped for use in children aged #24 with consensus reached by $1 round informative.
months, even if the age range of these of voting. The consensus statements
ASD-specific screening in children
screens exceeded 2 years (Table 1). A and discussion were summarized as
aged 18 to 24 months can assist in
PubMed search was conducted on June draft proceedings of the conference,
early detection
30, 2010, by using the search terms which were subsequently edited by all
participants. The search was updated Table 1 summarizes the measurement
(“child developmental disorders, perva-
by using the same strategy to add properties of ASD-specific level 1
sive” or “autistic disorder/” or “autism
articles published to December 31, screening tools for children aged ,36
[tw]” or “autistic [tw]”) and (“mass
2013, which yielded an additional 85 months. These include the following
screening” or “screen [tw]”), with the
references; selection was limited to tools.
age filter (“infant, birth-23 months”) and
limited to English-language articles. This prediagnostic screening of early be- CHAT
search yielded 111 references, which havioral or biological markers. The The CHAT was the first ASD screening
were reviewed by Drs Zwaigenbaum and working group reviewed and ap- tool to be assessed at a population
Bauman, who selected articles focusing proved the final wording of the sum- level.46 It cannot be recommended,
on studies that involved prediagnostic mary and recommendations. however, for current early detection
screening for early behavioral or The measurement properties that efforts due to its low sensitivity (18%,
biological features (as opposed to characterize the accuracy of screen- based on 6-year follow-up of a screened
postdiagnostic screening for etiologic ing instruments used to identify cohort of 18-month-olds).49
factors or associated comorbidities). children at risk for ASDs are sum- Q-CHAT
The search results were complemented marized in Table 1.17,20–47 ASD
The Q-CHAT extends the measurement
by additional publications identified by screeners with published evaluation
model of the CHAT, covering a broader
working group members. Thus, al- data include parent questionnaires
range of ASD symptoms, which are
though the search strategy was com- such as the Modified Checklist for
rated on a 5-point scale (rather than
prehensive, selection of articles was not Autism in Toddlers (M-CHAT),24 the present/absent). Preliminary data
systematic, which is an important limi- Quantitative Checklist for Autism in suggest that the Q-CHAT distinguishes
tation. A scoping approach was used Toddlers (Q-CHAT),33 the Early Screen- children with ASD from low-risk 18- to
instead, with some discretion of the ing of Autistic Traits questionnaire 24-month-olds.33 A recent secondary
multidisciplinary expert working group, (ESAT),22,23 and the First Year Inventory analysis using the 10 Q-CHAT items that
to select articles of highest relevance. (FYI).20,48 Table 1 also summarizes ASD best discriminated groups with and
Most of the instruments reviewed were screening instruments with only pre- without ASD and that optimized a
designed to identify children at risk for liminary data (eg, the Pervasive De- screening cut-point indicated sensitiv-
ASD who warranted further evaluation. velopmental Disorders Rating Scale),36 ity and specificity estimates as high as
Also reviewed were general develop- which will not be included in the 91% and 89%, respectively, in a case-
mental, or broadband, screening present discussion. control sample.34 Further validation of
instruments that had been evaluated The results of the overall process are this abbreviated screen is needed,
for the purpose of early identification of listed as summary statements. Some of however, in independent, community-
ASD, even if not specifically designed to the statements summarize the state of based samples similar to where the
distinguish risk for ASD from risk for the literature, whereas others provide screen would be used.
other developmental delays. We also recommendations for research needed
distinguished between the instruments to fill important evidence gaps and/or M-CHAT
that had been evaluated as level 1 address issues important for clinical The M-CHAT, also adapted from the CHAT,
screens, level 2 screens, or both. practice. has been assessed in large community
PEDIATRICS Volume 136, Supplement 1, October 2015 S43
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018TABLE 1 Parent-Report Screening Tools for Autism
S44
Screening Tool Reference Population (N, Age, Diagnosis, Level) Sensitivity and Specificity PPV and NPV Comments/Recommendation
ASD-specific screeners: parent report
FYI Reznick et al,20 2007 N = 698 infants aged 12 mo Preliminary findings: (PPV = 0.31 and NPV = 0.99) Promising tool for infants aged 12
(Turner-Brown General population mailing at 12 mo, N = 699 with outcomes at 42 mo. mo, but additional data needed
et al,21 2013) with follow-up at 42 mo Diagnosis with ASD = 9. FYI 2-domain
risk algorithm flagged 4/9 cases later
ZWAIGENBAUM et al
diagnosed with ASD
Sensitivity = 0.44; Specificity = 0.99
ESAT Dietz et al,22 2006; N = 31 724 from general population Sensitivity and specificity not reported PPV = 0.25 Not yet recommended as level 1
Swinkels et al,23 2006 Stage 1, n = 370 screened positive; Identified 18 ASD from 31 724 screened screener; additional data needed
Stage 2, of n = 255 100 screened positive
14–15 mo (mean: 14.9 mo)
M-CHAT Robins et al,24 2001 N = 1293, mix of low and high risk Estimates of sensitivity and specificity Strong evidence for use as both level
Mean: 14.9 mo (14–15 mo) cannot be determined from this study 1 and level 2 tool, 16–30 mo;
(screen-negative cases not additional data will be helpful,
systematically evaluated) especially in estimating
sensitivity
Robins,25 2008 N = 4797; 362 screened positive Estimates of sensitivity and specificity Without follow-up interview, PPV = 0.058
(qualified for follow-up interview); cannot be determined from this study
16–26 mo (screen-negative cases not
15-, 18-, and 24-mo well-child visit results systematically evaluated) With follow-up interview, PPV = 0.57
If screen-positive cases are examined
for any significant developmental
delay, the PPV for M-CHAT + follow-up
interview is .0.90 across studies
Kleinman et al,26 2008 n = 3309 low risk, n = 484 high risk Estimates of sensitivity and specificity PPV = 0.11 for low-risk sample, 0.60 for
16–30 mo cannot be determined from this study high-risk sample, without interview.
(screen-negative cases not This improved to 0.65 (low risk) and
systematically evaluated) 0.76 (high risk) when follow-up
interview was considered part of the
screening procedure
Pandey et al,27 2008 n = 6050 low risk and n = 726 high risk Estimates of sensitivity and specificity PPV = 0.43 for low-risk samples (younger
cannot be determined from this study and older combined for this table) and
(screen-negative cases not 0.76 for high-risk samples; PPV
systematically evaluated) calculated based on M-CHAT + follow-
up interview
16–30 mo Note: Sample overlaps with Kleinman
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018
et al,26 2008, but does not include
samples from Robins et al,24 2001, or
Robins,25 2008
Inada et al,28 2011 N = 659; 18 mo Estimates of sensitivity and specificity PPV = 0.733
cannot be determined from this study
(screen negative cases not
systematically evaluated)TABLE 1 Continued
Screening Tool Reference Population (N, Age, Diagnosis, Level) Sensitivity and Specificity PPV and NPV Comments/Recommendation
29
Canal-Bedia et al, 2011 Validity study: 2417 low risk and 63 high Estimates of sensitivity and specificity Validity study: PPV not reported In validity study, 19 of 23 children
risk; 18–36 mo cannot be determined from this study separately for low-risk and high-risk diagnosed with ASD were from
(screen negative cases not samples high-risk sample.
Reliability study: 2055 low risk; 18–36 systematically evaluated) Reliability study: PPV = 0.19 In reliability study, rate of ASD in low-
mo risk sample was 2.9 in 1000
Pinto-Martin et al,30 2008 N = 152; 18–30 mo No diagnostic follow-up, cannot assess
psychometrics; comparison of M-CHAT
and PEDS
Chlebowski et al,31 2013 N = 18 989, 18–30 mo Estimates of sensitivity and specificity Among screen-positive children who Emphasizes potential clinical utility
cannot be determined from this study were evaluated (171 of 278 [60.7%]) of M-CHAT as level 1 screen (high
PPV)
Note: some potential false-negative PPV = 0.538 for ASD (if any DD is included, Authors suggested that if initial M-
findings ascertained by concurrent PPV increases to 0.977) CHAT score is $7, the follow-up M-
PEDIATRICS Volume 136, Supplement 1, October 2015
screening using other instruments Note: Sample overlaps with Kleinman CHAT interview may not be
et al,26 2008; Pandey et al,27 2008; needed due to high PPV for ASD
Robins et al,24 2001; Robins et al,25 (.0.80). However, the follow-up
2008 M-CHAT interview is essential for
children with initial scores of 3–6
M-CHAT-R/F Robins et al,32 2014 N = 16 115 low-risk toddlers Estimates of sensitivity and specificity Among screen-positive children who Children with ,3 items endorsed
cannot be determined from this study were evaluated (221 of 348 [63.5%]) (93% of all cases) did not require
Note: some potential false-negative PPV = 0.475 for ASD (if any DD is included, the follow-up interview or any
findings ascertained by concurrent PPV increases to 0.946) other evaluation. Children with 3–
screening using other instruments 7 items endorsed (6% of all
cases) required the follow-up
interview; if at least 2 items
remained positive, then referral
for diagnostic evaluation was
indicated. Children with $8 items
endorsed (1% of all cases) were
at sufficiently high risk to be
referred directly for diagnostic
assessment. Using this strategy
reduced the case positive rate
(from 9.2% to 7.2%) without
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018
significant change to PPV, relative
to previous follow-up M-CHAT
strategy
Q-CHAT Allison et al,33 2008 779 low-risk toddlers with mean age of 21 Sensitivity and specificity not provided Not reported
mo; plus
160 toddlers and preschoolers with ASD
with mean age of 44 mo
S45
SUPPLEMENT ARTICLETABLE 1 Continued
S46
Screening Tool Reference Population (N, Age, Diagnosis, Level) Sensitivity and Specificity PPV and NPV Comments/Recommendation
34
Allison et al, 2012 754 controls; mean: 36 mo (drawn from Sensitivity = 0.91; Specificity = 0.89 PPV = 0.58 (with pretest odds = 0.16 Clinical diagnoses based on parent
low-risk sample in Allison et al,33 2008) based on available sample) report, with recruitment through
Web-based research registry
126 toddlers and preschoolers with ASD Based on screening cut-point of 3 from Further evaluation in independent
(aged 15–47 mo; mean: 20.8 mo) derivation sample, using the 10 of 25 samples warranted
ZWAIGENBAUM et al
Randomly allocated to derivation and items from original Q-CHAT that best
validation samples discriminate groups
DBC-ES Gray et al,35 2008 N = 207; 20–51 mo; level 2 Sensitivity = 0.83 (estimated); Specificity
= 0.48 (estimated)
PDDRS Eaves and N = 199 with autistic disorder, rated by Factor analysis, no psychometrics Insufficient data to evaluate utility as
Williams,36 2006 teachers, teaching interns, and family calculated screening tool for young children
members; aged 1–6 y
Eaves et al,37 2006 N = 134, rated by teachers, teaching Autistic disorder, cutoff 85: Sensitivity = Not reported
interns, or parents; aged 3–26 y 0.93 and specificity = 0.48
(mean: 9.7 y); diagnosis: autism (n = Autistic disorder, cutoff 90: Sensitivity =
86), Asperger disorder (n = 11), PDD- 0.84 and specificity = 0.58
NOS (n = 15), non–ASD disorder (n = PDD, cutoff 85: Sensitivity = 0.88 and
23) specificity = 0.68
PDD, cutoff 90: Sensitivity = 0.78 and
specificity = 0.77
ASD-specific screeners: interactive observational measures
STAT Stone et al,38 2000 n = 40 (development sample), n = 33 Sensitivity = 0.83 and specificity = 0.86 for Not reported Strong evidence for use as level 2
(validation sample); 24–35 mo, level 2 validation sample (sensitivity = 0.83 tool, 24–35 mo; promising for 14–
(high risk) and specificity = 0.83 for development 23 mo but additional data will be
age-matched subsample) helpful
Stone et al,39 2004 Study 1: N = 52, 24–35 mo, ASD and other Study 1: one-half of sample used to Study 1: PPV = 0.86 and NPV = 0.92
developmental delay matched on determine cutoff with optimal (validation subsample)
chronological and mental age, level 2 sensitivity/specificity and one-half
used to validate cutoff of 2: Sensitivity
= 0.92 and specificity = 0.85
Study 2: N = 104, 24–35 mo, level 2 Study 2: not reported, but based on table
provided, sensitivity = 1.0 and
specificity = 0.90 for autistic disorder
(lower for PDD-NOS)
Stone et al,40 2008 N = 71, 12–23 mo, level 2 (follow-up Cutoff of 2: Sensitivity = 1.0 and specificity Cutoff of 2: PPV = 0.38 and NPV = 1.0
assessment 24–42 mo) = 0.40
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018
Cutoff of 2.75: Sensitivity = 0.95 and Cutoff of 2.75: PPV = 0.56 and NPV = 0.97
specificity = 0.73
Cutoff of 2.75 in subsample of children Cutoff of 2.75 in subsample of children
14–23 mo (n = 50): Sensitivity = 0.93 14–23 mo (n = 50): PPV = 0.68 and
and specificity = 0.83 NPV = 0.97TABLE 1 Continued
Screening Tool Reference Population (N, Age, Diagnosis, Level) Sensitivity and Specificity PPV and NPV Comments/Recommendation
41,42
BISCUIT Matson et al, 2009 N = 1007 sample with ASD or DD aged 330 confirmed ASD diagnosis Promising as diagnostic tool or level
17–37 mo (mean: 26.4 mo) Sensitivity = 84.7; Specificity = 86.4 2 screener; more data needed
AUC = 0.55
SORF Wetherby et al,43 2004 N = 150, level 2 screen of low-risk 20 Significant red flags; cutoff of 8 red Not reported Promising as level 2 screener;
sample of 6581 children aged 18–24 flags reported data based on coding
mo Sensitivity = 0.87 and Specificity = 0.84 from video rather than office-
AUC = 0.93 based assessment; more data
needed
Broadband screener: parent report
CSBS ITC Wetherby et al,43 2004; N = 5385 children aged 6–24 mo 60 ASD from 5385 screened Note: PPV and NPV are not reported for Need additional data, but promising
Wetherby et al,44 2008 screened; follow-up evaluation of n = ASD specifically but for all delays tool for 9–.24 mo; not
813; general population (consecutive collapsed into 1 group recommended for age ,9 mo
screens from 6 to 24 mo; diagnostic Mean: 16.4 mo (6–24 mo) PPV and NPV for communication delay Note: ITC positive screen does not
PEDIATRICS Volume 136, Supplement 1, October 2015
outcome/follow-up questionnaire at 4 vary with age: distinguish communication delay
+ years) Sensitivity = 88.9–94.4; Specificity = 88.9 PPV = 0.42–0.79 from ASD; second-level ASD-
Sensitivity estimated at 0.93 (56/60 NPV = 0.87–0.99 specific screen recommended
identified when ,6 consecutive
screens were done across 18-mo
period)
Pierce et al,17 2011 N = 10 479 infants whose parents 32 with ASD from 10 479 screened PPV = 0.75 (estimated) for developmental Promising tool; further study needed
completed checklist at 1-y well-child Sensitivity and specificity not provided delays including ASD to examine sensitivity and
visit (mean age: 12.54 mo); level 1 specificity in general pediatric
screen settings
Oosterling et al,45 2009 N = 238; 8–44 mo (mean: 29.6 mo) See note (at right) See note (at right) Note: True sensitivity, specificity, PPV,
Overall sensitivity reported at 0.71 Overall PPV reported at 0.78 and NPV cannot be estimated
Overall specificity reported at 0.59 Overall NPV reported at 0.50 because another tool (ESAT) was
Because norms for ITC are available only used as “prescreen” and only
for children aged 6–24 mo, it is ESAT screen-positive cases were
difficult to interpret use as a screener subsequently screened with ITC
for children aged .24 mo
Adapted from the table developed by the Autism Subcommittee for the National Children’s Study. Other instruments not listed in the table: CHAT (Baron-Cohen et al,46 1992; Baron-Cohen et al,47 1996; and other articles); longitudinal data indicated poor
sensitivity; not recommended for use as level 1 screen; and Social Responsiveness Scale–Preschool (SRS-P), under development for screening preschool-aged children. ADI-R, Autism Diagnostic Interview–Revised; AOSI, Autism Observation Scale for
Infants; BISCUIT, Baby and Infant Screen for Children with Autism Traits; CSBS ITC, Communication and Symbolic Behavior Scales Infant Toddler Checklist (or Infant Toddler Checklist, ITC); DBC-ES, Developmental Behavior Checklist–Early Screen; DD,
developmental delay; ESAC, Early Screening for Autism and Communication Disorders; ESAT, Early Screening of Autistic Traits; M-CHAT-R/F, M-CHAT, Revised With Follow-Up; PDD-NOS, pervasive developmental disorder not otherwise specified; PDDRS,
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018
Pervasive Developmental Disorder Rating Scale; Q-CHAT, Quantitative - Checklist for Autism in Toddlers; SORF, Systematic Observation of Red Flags of ASD; STAT, Screening Tool for Autism in Toddlers & Young Children.
S47
SUPPLEMENT ARTICLEsamples as a level 1 screen. The 23-item 1000) (Table 1). Another study that Early Screening for Autistic Traits
M-CHAT questionnaire, combined with evaluated the psychometric proper- Population screening at an even earlier
a follow-up interview to help clarify ties of the Spanish version of the age has been associated with higher
items endorsed by parents on the M-CHAT in a community sample of false-negative rates (lower sensitivity),
initial screen, is estimated to have children in Mexico reported similar which is somewhat expected given the
a PPV as high as 0.57 to 0.65 in low-risk discriminative validity, 50 although slow onset of symptoms that emerges
samples.25,26,31 Pandey et al27 reported some items appeared less informative across the first 24 months of life. The
that the PPV of the M-CHAT (as used for for ASD than in published reports on ESAT was assessed in a large (N = 31
first-level screening in a low-risk the original English-language version. 724) population sample of 14- to 15-
community sample with follow-up in- Psychometric data on Japanese28 and month-olds, with a low case detection
terview) is lower in younger children, Arabic51 translations have also been rate (,1 in 1000).22,23 Moreover, PPV of
with a PPV of 0.28 in toddlers aged 16 reported. (Additional information on the ESAT was only 0.25, which would
to 23 months compared with a PPV of available translations of the M-CHAT potentially lead to the referral of
0.61 in those aged 24 to 30 months. is available at http://www2.gsu.edu/ a large number of toddlers without ASD
There are many reasons for false- ∼psydlr/Site/Official_M-CHAT_Website. based on a positive screen (PPV for
positive findings, including develop- html [accessed October 17, 2014]). other developmental delays was not
mental concerns that may resolve and reported). The authors recommended
Recently, Robins et al32 reported vali-
behaviors in typically developing tod- a second screening at 24 months of age
dation data for a new version of this to identify children who regress after
dlers that overlap with ASD deficits,
screening tool, the M-CHAT, Revised age 18 months or those who are
such as repetitive behaviors (eg,
with Follow-Up, in 16 115 toddlers. The missed for other reasons.
turning lights on and off) and re-
stricted interests (eg, insistence on questionnaire was reduced to 20
routines).19 However, despite lower items, removing 3 items that had Baby and Infant Screen for Children
specificity for autism at 18 months, performed poorly (“peek-a-boo,” With Autism Traits
PPV for any diagnosable develop- “playing with toys,” and “wandering Preliminary data on the Baby and Infant
mental disorder was high for all without purpose”); wording on other Screen for Children with Autism Traits
groups. In the largest sample of tod- items was simplified and/or examples tool indicate good discrimination
dlers (aged 18–30 months) reported provided for further clarity. A scoring between toddlers with known ASD
to date (N = 18 989 [including some algorithm with 3 risk ranges was de- diagnoses and those with other de-
children in previous reports]),25–27 the veloped. Children in the low-risk range velopmental delays as identified clini-
PPV of the M-CHAT for ASD was 0.54, (ie, ,3 items endorsed) did not re- cally.41 Additional data are needed,
and for any developmental disorder, it quire the follow-up interview or any however, to confirm how this measure
was 0.98.31 As in other community- other additional evaluation (93% of all would perform in a screening context.
based ASD-screening studies, esti- cases). Children in the medium-risk
mates of PPV were based on those range (ie, 3–7 items endorsed [6% of FYI
screen-positive children who attended all cases]) required the follow-up in-
The FYI is a parent questionnaire
and completed a diagnostic evaluation terview to clarify their risk for ASD; if at
designed to screen for signs of autism in
(39.3% of screen-positive children least 2 items remained positive, then
12-month-olds. Initial data on the FYI
were not assessed). referral for diagnostic evaluation was
suggest the potential for modest sensi-
The M-CHAT has also been evaluated indicated. Children in the high-risk range tivity.20 In a recent prospective follow-up
internationally and in multiple lan- (ie, $8 items endorsed [1% of all cases]) study of a community sample of 699
guages. Canal-Bedia et al29 assessed were at sufficiently high risk to be re- children whose parents initially com-
the reliability and predictive validity of ferred directly for diagnostic assess- pleted the FYI at approximately the
a Spanish translation of the M-CHAT in ment without the follow-up interview. child’s first birthday, 4 of 9 children
a combined community and at-risk This revised scoring and referral algo- subsequently diagnosed with ASD at 3
sample in Spain. The PPV in the com- rithm reduced the initial screen-positive years of age were identified. A scoring
munity sample was 0.19, although this rate (from 9.2% to 7.2%) and increased algorithm that optimized prediction of
finding may have reflected a relatively the overall rate of ASD detection (67 vs 45 ASD identified 13 (1.9%) of 699 partic-
low base rate of identified preschool- per 10 000) compared with the original ipants who met cutoffs on 2 domains
aged children with the disorder (2.9 in follow-up M-CHAT. (social communication and sensory
S48 ZWAIGENBAUM et al
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018SUPPLEMENT ARTICLE
regulation).21 Assessment of PPV in an Systematic Observation for Red Flags sions); 184 ultimately received further
independent/validation sample is still The Systematic Observation for Red evaluation. Of this group, 32 toddlers re-
needed. Flags has shown promise in discrimi- ceived an ASD diagnosis by age 3 years.
The working group suggested that ad- nating ASD from other communication This general population screening ap-
ditional efforts are needed to develop delays.43 Additional data are needed in proach also detected 65 toddlers with
and validate population-based ASD a screening context. a language delay or global developmental
screening tools aimed at the 12- to 18- delay, and 36 children with other delays.
Broadband screening in children Thus, the PPV for detecting toddlers with
month age range, anticipating that
aged ,24 months can assist in early ASD or developmental delay in this study
modest sensitivity at this age may war-
detection of ASD was estimated to be 0.75. Importantly, all
rant follow-up with additional screening
at a later age (eg, at 24 months). In ad- Delays and deviances in social commu- toddlers identified with delays were re-
dition, the working group recommended nication are often subtly present around ferred for treatment, and the majority
that standardized screening specifically thefirstbirthdaybutareoftennotstrongly started intervention well before their
for ASD should be performed when ASD-specific at that early age. Broadband second birthday.
parents raise concerns between well- developmental screening tools, such as This research illustrates that autism can
child visits or when concerns are the Communication and Symbolic Be- sometimes be detected by the first
raised upon general developmental havior Scales Developmental Profile birthday by using a broadband de-
surveillance or screening during (CSBS DP) Infant/Toddler Checklist de- velopmental screen in real-world pedi-
scheduled visits. Parental concern ef- veloped by Wetherby and Prizant,53 were atric practices as standard of care. The
fectively raises the prior probability that shown to be effective at detecting autism CSBS DP Infant/Toddler Checklist is not
a child will have ASD, thereby increasing before the onset of full-blown clinical specific for ASD (ie, does not differentiate
the PPV of a screening test regardless of symptoms. Wetherby et al44 evaluated ASD from other communication dis-
its intrinsic sensitivity and specificity. the CSBS DP Infant/Toddler Checklist in orders), but follow-up evaluation by
a community sample of 5385 children a developmental specialist (eg, speech
Level 2 Screening Tools aged 6 to 24 months recruited from language pathologist, psychologist, de-
Two interactive observational assess- health and child care services. The velopmental behavioral pediatrician)
ments have been developed for use as Infant/Toddler Checklist identified 56 can help determine the need for ASD-
level 2 screeners in young children (93%) of 60 children with ASD classified specific diagnostic assessment as well
identified as being at high risk of ASD. independently at age 3 years in a con- as identify other developmental delays in
current prevalence study of the same need of support and intervention. Use
region. Some Infant/Toddler Checklist of even broader, more general de-
Screening Tool for Autism in findings were positive as early as 9 to 11 velopmental screening tools, such as the
Two-Year-Olds months, although in some cases, an ini- Parents’ Evaluation of Developmental
The Screening Tool for Autism in Two- tial screen was negative at 9 to 11 Status (PEDS)54,55 and the Ages & Stages
Year-Olds (STAT) has been assessed in months and did not become positive until Questionnaire,56 to detect ASD are under
clinical samples of 2-year-olds referred a later administration. The Infant/Toddler investigation. Because these tools are
for suspected ASD, with a sensitivity and Checklist also identified concerns sooner commonly used in pediatric practice, it
specificity as high as 92% and 85%, re- and more consistently than an open- will be important to determine their
spectively.39 Recent data indicate that ended question about parents’ de- utility in detecting ASD in the second
the STAT may also have utility in younger velopmental concerns. Subsequently, year of life even though their sensitivity
toddlers aged 14 to 23 months, although Pierce et al17 assembled a network of and specificity are not expected to be as
additional data are needed for this age 137 pediatricians who administered the high as those of ASD-specific screeners.
group.40 Although the STAT requires CSBS DP Infant/Toddler Checklist at ev-
a higher level of expertise to administer ery routine 1-year check-up examination. Statement 2: The evidence
than parent questionnaires such as the Of ∼10 000 screens administered, 1318 indicates that siblings of children
M-CHAT, a recent study provided evi- children failed the screen. The pedia- with ASD are at elevated risk for
dence of the effectiveness of Web-based tricians referred 346 screen-positive ASD and other developmental
training of community services pro- children as “at-risk” children (the disorders and thus should receive
viders of various professional back- screening was thus embedded within intensified surveillance.
grounds; this training could enhance a surveillance context, in which clinical Based on data from a US register of 2920
the feasibility of the STAT.52 judgment contributed to referral deci- children aged 4 to 18 years in families
PEDIATRICS Volume 136, Supplement 1, October 2015 S49
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018affected by ASD, the frequency of ASD in suggest that such benefits of early although only 1 study focused almost
a later-born sibling has been estimated at screening frequently go unrealized. In a exclusively on this age group.67 These
14%.57 More recently, several independent national study of 17 pediatric practices, studies provide promising evidence of
groups conducting prospective longitudi- implementation of general develop- the stability of ASD diagnosed as early as
nal research involving infant siblings of mental screening did not always lead to 14 months; the samples were relatively
children with ASD reported a pooled es- referral of screen-positive children to small, however, and there is no direct
timated recurrence risk of 18%.58 In con- a medical subspecialist or early in- comparison of stability in children di-
trast, a recent population registry-based tervention programs.12 These inves- agnosed before versus after age 24
study from Denmark59 estimated re- tigators noted that some families did months. Of note, 2 studies focused on
currence risk at closer to the 7% to 8% not understand the reason for a follow- toddlers identified by using community-
level reported in older studies.60 Regard- up evaluation. Additional research is level ASD screening before age 24
less, rates of ASD in siblings greatly ex- needed to address how to better engage months.65,66 Both studies indicated high
ceed population risk, emphasizing the families in the screening process to fa- diagnostic stability for children initially
need for intensified monitoring. More- cilitate rapid follow-up, as well as to diagnosed with autistic disorder (85%–
over, younger siblings of children with identify and characterize other potential 93%) but more modest stability for
ASD demonstrate significant deficits on barriers to early diagnosis and treat- children diagnosed with pervasive de-
indices of social communicative develop- ment related to system capacity or velopmental disorder not otherwise
ment and cognitive functioning, as well as provider attitudes and practices. specified (47%–62%). Further research
elevated ASD symptoms relative to youn- in larger samples is needed, but the
ger siblings of typically developing chil- Statement 4: The long-term evidence to date supports the stability of
dren.61–64 Because these children are at stability of ASD diagnosis in ASD diagnoses before age 2 years.
elevated risk, they require intensified children aged ‡24 months is well
developmental surveillance. At a mini- established. Emerging data Statement 5: Further attention to
suggest that ASD diagnoses in potential barriers to ASD-specific
mum, they should receive continuous
substantial proportions of children screening in the health care
surveillance for developmental issues
diagnosed before age 24 months system is needed.
and be screened for ASD at 18 and 24
are also stable, although further
months of age, as recommended by the Pediatricians have noted major barriers
research is needed, particularly in
AAP for all children.2 the context of early screening. to screening, including the following: lack
of time and inadequate reimbursement;
Statement 3: Children identified Ten articles were identified in which
logistic challenges, such as disruption of
through ASD-specific screening children received an initial diagnostic
work flow, lack of familiarity with tools,
should be immediately referred for assessment for possible ASD before age
diagnostic/developmental and difficulty with scoring; and lack of
3 years and were then reassessed at
evaluation and appropriate office-based systems for making refer-
least 1 year later.65–74 In general, the
intervention. rals and monitoring outcomes.
stability of ASD diagnoses established
The AAP has recommended that children at $24 months (ie, the rate at which Lack of Time and Reimbursement
who screen positive on an ASD-specific an ASD diagnosis was confirmed on Insufficient time and inadequate re-
screening tool be scheduled for a com- reassessment) was very high, ranging imbursement are often cited by pro-
prehensive evaluation and referred from 68.4% to 100% when the initial viders as barriers to performing
concurrently to early intervention ser- diagnosis was autistic disorder (me- screening.12,13,19,75 Pediatricians have
vices as appropriate.2 Available inter- dian: 92%), and from 40% to 100% a limited amount of time to complete an
ventions are mandated in the United when the initial diagnosis was perva- increasing number of tasks, including
States but vary in availability and quality sive developmental disorder not oth- screening for non-ASD disorders, during
by locality, and they may consist of non– erwise specified (according to the a well-child visit.19 Selection of a broad-
ASD-specific public early intervention Diagnostic and Statistical Manual of band screening instrument would meet
programs, such as speech therapy, and Mental Disorders, Fourth Edition, or with greater acceptance if the tool could
early childhood education programs. the Diagnostic and Statistical Manual detect multiple developmental dis-
It is hoped that early screening will lead of Mental Disorders, Fourth Edition, orders of interest. Busy periods, such as
to improved outcomes as a result of Text Revision [median: 61%]). the onset of the winter viral season, of-
earlier referral and earlier initiation of Four of these studies involved samples of ten impede the ability of a practice
intervention. However, recent studies children aged ,24 months (Table 2),65–68 to consistently screen.12 To optimize
S50 ZWAIGENBAUM et al
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018SUPPLEMENT ARTICLE
TABLE 2 Studies of Diagnostic Stability That Include Children Initially Assessed With ASD Before 2 Years of Age
Reference Sample Mean Age, Age Mean Age, Age Diagnosis N Diagnosis at T2 N % Stability
Range at T1, mo Range at T2, mo at T1
van Daalen et al,65 2009 Population-based sample 23 43 (34–64) Autism 40 ASD 38 95.0
Non-ASD 2
PDD-NOS 13 ASD 8 61.5
Non-ASD 5
Non-ASD 78 ASD 2 97.4
Non-ASD 76
Kleinman et al,66 2008 Mixed level 1 (physician office) 26.7 (16–35) 52.9 (41–82) Autism 46 ASD 39 84.8
and level 2 (early intervention, Non-ASD 7
sibling) sample PDD-NOS 15 ASD 7 46.7
Non-ASD 8
Non-ASD 16 ASD 0 100
Non-ASD 16
Chawarska et al,67 2007 Referrals to specialty clinic 21.6 (14–25) 35.9 Autism 21 ASD 21 100
with suspected ASD Non-ASD 0
PDD-NOS 6 ASD 6 100
Non-ASD 0
Non-ASD 4 ASD 1 75
Non-ASD 3
Gillberg et al,68 1990a Referred sample 23.0 (8–35) 57.7 (36–140) Autism 21 ASD 21 100
Non-ASD 0
PDD-NOS 4 ASD 2 50
Non-ASD 2
Non-ASD 2 ASD 0 100
Non-ASD 2
PDD-NOS, pervasive developmental disorder not otherwise specified; T1, initial diagnostic assessment of ASD; T2, reassessment of ASD diagnosis, at least 1 year later in these studies.
a One child, diagnosed at 8 months, was followed up only to age 26 months and thus was excluded from the table.
screening, some practices have in- ruption of work flow, unfamiliarity with were referred, and many practices
stituted ongoing data collection and screening instruments, and difficulty with struggled to track their referrals.12
monitoring of their efforts. scoring.12,13,75 Providers often express Practice-specific referral rates varied
The lack of reimbursement for screening concerns about how to distribute widely, from 27% to 100%. It is important
iscommonlycitedasabarrier.However,in screening questionnaires without slowing that each pediatric practice establish
1 study, the 3 practices that routinely the flow of patients through the office.12,13 a specific implementation system to ex-
screened at the 30-month well-child visit Nevertheless, in a national sample of 17 pedite referrals, communicate with spe-
reported no difficulties in collecting pay- pediatric practices, .85% of children cialists and early intervention programs,
ment.12 In another study,17 pediatric offi- presenting at recommended screening and track follow-through and outcomes.
ces received no payment at all for ages were screened, with practices di- Clearly, early screening initiatives are
screening but rather received training viding responsibilities among staff only as effective as access to resources
and data collection support, as well as members and proactively monitoring for follow-up evaluation and early in-
streamlined follow-up diagnostic assess- implementation.12 Miller et al76 found that tervention. Communication back to the
ments for screen-positive children. Thus, screening at sick visits was necessary to referring office relative to the outcomes
reimbursement challenges may be me- achieve coverage of the age-eligible chil- of follow-up actions is critical if only to
diated by infrastructure support (eg, staff dren, especially for the small number of reassure all concerned of the value of
training/mentoring) to make screening uninsured children. Training of office staff such referrals. For children with ASDs,
easier to implement, as well as timely as well as professional education can early intervention services have become
access to appropriate follow-up. In this remedy a lack of familiarity with the use more accessible through Part C of the
way, pediatricians may be reassured that and scoring of screening tools. 2009 Individuals With Disabilities Educa-
there is capacity in the health system to tion Act but access may not be equal in
Lack of Office-Based Systems for all parts of the country, and the quality of
support children who screen positive.
Making Referrals and Monitoring services can vary widely and affect out-
Logistic Challenges Outcomes comes. Indeed, although the National
Other challenges to screening imple- In the sample of 17 pediatric practices, Research Council has recommended
mentation include concerns over a dis- only 61% of children with failed screens entry into an intervention program as
PEDIATRICS Volume 136, Supplement 1, October 2015 S51
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018soon as an ASD is suspected,3 local up tracking of screen-negative in a complementary fashion to max-
factors, including funding, can affect cases, to improve estimates of sen- imize both sensitivity and specificity
access to services (wait-listing) or make sitivity, specificity, and NPV of early screening, perhaps in the
certain early intervention programs evaluation of different scoring context of multistage screening, in
unavailable to some children.75,77 approaches (categorical versus which a wide net is cast initially
Thus, barriers to screening can be over- continuous) and, potentially, differ- and false-positives are winnowed
come with specific strategies such as ent age-specific scoring algorithms out in successive assessments
training and involvement of clinic staff and for specific ages, to further optimize Evaluate screening strategies by us-
use of reminder systems, even in busy screening strategies that might be ing randomized experimental designs
practices. However, better-coordinated implemented longitudinally Consider additional outcome met-
efforts are needed to ensure access to reporting of detailed characteriza- rics for screening: potential finan-
specialized assessment and intervention tions of study participants, includ- cial savings to society, unintended
for children at risk identified through the ing social factors, cognitive level, effects (eg, family stress)
screening process, as well as communi- and medical history, to improve Examine whether computer technol-
cation back to community pediatricians. In comparisons across studies and ogy can improve screening accuracy
addition, further consideration is needed to better understand what factors
regarding how physician beliefs related to Examine the effectiveness of re-
might influence the accuracy of
ASD screening (eg, potential risks and peated screening for ASD
screening for individual children
Evaluate how belief systems affect
benefits to children and families, system evaluation of potential differences screening uptake and outcomes
capacity to provide timely specialized as- between screen-positive children
sessment and treatment services) may who are seen for a diagnostic as- Examine potential screening strat-
influence practice behavior. Such beliefs sessment and those who do not egies that include measurement of
can contribute to incongruence between complete follow-up (which is often biomarkers
physician knowledge and actions when in the range of 25%–40%25,27 and in
managing ASD-related concerns78 and some studies exceeds 50%17) to Examine how broadband and
thus may also need to be addressed to further evaluate potential barriers ASD-specific screening tools can be
facilitate uptake of ASD screening into and facilitators, and provide infor- used in a complementary fashion to
community pediatric practices. mation essential to evaluating the maximize both sensitivity and
Statement 6: Methodologically generalizability of study findings specificity of early screening
rigorous research in ASD-specific inclusion of underrepresented mi- Can a general developmental tool be
screening should be a high priority. nority and historically underserved relied on to identify children who should
Future research in ASD screening would groups, to help ensure representa- be evaluated for ASD? If a broadband
be aided by attention to the following tive samples and the development of screening tool is indeed dependable, as
methodologic issues: culturally appropriate adaptations of suggested by Wetherby et al43,44 and
screening tools for such populations Pierce et al,17 then a multistage screen-
use of large, representative high- ing strategy focusing on routine sur-
and low-risk samples, to strengthen Lower socioeconomic status and non-
white ethnicity (particularly Hispanic) veillance and use of a broadband
the generalizability of findings
have been associated with delayed age screening tool, followed by an ASD-
use of meaningful end points (eg, of diagnosis, potentially due to dis- specific instrument for children who
validated diagnostic measures to test positive on the initial screen, can
parities in access to health services.79–81
assess for ASD and other develop- help reduce the need for extra testing
However, there is evidence that appli-
mental disorders, as well as an and the additional clinic time and effort.
cation of standardized screening can
increased focus on outcomes of help reduce such disparities and en- A notable value of this approach is the
greatest relevance to families and sure timely diagnosis of children across limiting of referrals for specialized as-
to the health system, such as age of a diversity of backgrounds.82 sessment, without sacrificing case de-
diagnosis, age of entry into inter- tection rate. If broadband screening
vention, and long-term developmen- Statement 7: Additional priorities for cannot reliably detect ASD, then a
tal gains resulting from screening) future research include studies that: screening strategy mandating ASD-
inclusion of systematic surveil- Examine how broadband and ASD- specific screening for all children,
lance methods, as well as follow- specific screening tools can be used alongside broadband screening to
S52 ZWAIGENBAUM et al
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018SUPPLEMENT ARTICLE
detect other potential developmental to screening (eg, broadband versus ASD- distal outcome. For nonfatal conditions,
concerns, would be more appropriate. The specific, level 1 versus level 2, or some developing approaches to measure im-
first approach was described by Filipek combination) have never been directly pact on morbidity, disability, or impair-
et al83; the second approach is currently compared. We would argue that screen- ment can be a challenge. With respect to
recommended by the AAP.2 Unfortunately, ing is a public health intervention; that is, ASD, although increases in referral and
the effectiveness (and cost-effectiveness) a comprehensive early detection strategy early diagnosis rates can serve as
of the 2 strategies has not been well shouldnotbesolelybasedontheselection meaningful initial outcomes, screening
studied. Data from a single pediatric of a particular screening instrument but should ultimately demonstrate a re-
practice showed that ∼75% of children rather must include other changes to the duction in population impairment and the
with positive results on the ASD-specific overall system of care, such as enhanced effect of that impairment on society.
screening tool (the M-CHAT) were missed training for health professionals and ex- Studies of ASD screening will thus even-
by the PEDS, a standardized general de- panded capacity for early diagnosis and tually need to consider the impact of this
velopmental screening questionnaire.30 It intervention by specialized teams. Thus, screening on long-term changes in
should be noted, however, that this study the outcomes of screening may not simply symptoms and functional status. De-
did not report actual ASD diagnoses but berelatedtothemeasurementproperties termining how to best measure these
rather simply examined agreement in of a tool but also to the successful distal health outcomes is one of the
screening classification by the 2 tools. implementation of other aspects to the challenges of ASD research. In addition to
However, Wiggins et al54 reported that the overall care pathway for children with distal health outcomes, assessing the
M-CHAT had higher sensitivity for ASD than suspected ASD.17,84 As such, researchers cost impact of screening is often critical
the high-risk threshold for any area of should explicitly define their screening to its eventual broad dissemination.
general concern covered by the PEDS. Al- strategy (ie, the screening instrument Because ASDs impose a sizable financial
though the PEDS detected many children plus collateral changes to the system of burden, not only in direct medical expen-
with other developmental concerns, sen- care) as well as the outcomes of interest, dituresbutalsoinindirectcosts(eg,special
sitivity for ASD could not be achieved and evaluate the effectiveness of these education services, lost productivity by
without lowering the screen-positive strategies in real-life community settings family caregivers),87–89 a more in-depth
threshold to a level that would identify by using randomized designs. Random- understanding of these costs is needed
a substantial proportion of the general ized designs have become the standard to adequately compare different screening
population (25%). in other ASD intervention research (eg, strategies and to identify potential cost
A study assessing the efficacy of such Dawson et al5) and in other public health savings to society for those that are ef-
a multistage screening program would screening interventions.85 However, ob- fective. Finally, indirect costs associated
also assess/validate the effectiveness servational studies will also need to be with screening include an emotional di-
of: (1) training of health care pro- continued because of the well-known mension. Evaluations of screening effec-
fessionals in recognizing early ASD challenges to constructing randomized tiveness, in addition to including distal
signs and using a specific screening designs that reflect real-world clinical outcomes, need to consider these “costs”
tool; (2) a specific referral protocol; and practice.86 Table 3 presents a comparison in addition to the financial costs associ-
(3) feedback to the referring offices. of the relative strengths and limitations ated with false-positive findings.
of randomized and observational designs
with respect to screening research.
Evaluate screening strategies by Examine whether computer
using randomized designs technology can improve screening
The evaluation of ASD screening is often Consider additional outcome metrics accuracy
limited to measurement of classification for ASD screening The use of computer technology holds
accuracy (estimates of sensitivity and In the near term, evaluation of ASD promise for improving screening accu-
specificity, and/or PPV and NPV) without screening strategies will likely continue racy. Parents can complete a screening
sufficient attention to whether the ulti- to focus on process measures, such as questionnaire online and have access
mate goals of screening are achieved rates of targeted children screened, re- to video exemplars for more accurate
(eg, earlier diagnosis and access to ferred, and diagnosed. However, ulti- reporting. The capability to upload vid-
treatment) or the possibility that, as with mately, the idea of evaluating any eos can expedite specialist evaluations.
other interventions, screening might be screening program is to gauge its impact A recent preliminary report suggested
associated with positive or adverse out- on distal health outcomes. For potentially that the M-CHAT (including follow-up
comes. Moreover, alternate approaches fatal conditions, mortality is the ultimate questions) could be feasibly completed
PEDIATRICS Volume 136, Supplement 1, October 2015 S53
Downloaded from http://pediatrics.aappublications.org/ by guest on April 13, 2018You can also read
