Communication Interventions for Minimally Verbal Children ... Kaiser et al 2014...sessments. Of the children, 60 met ADOS20 criteria for autism, and 1 met a criterion for ASD. Participants

NEW RESEARCH

JOURNAL

VOLUM

Communication Interventions for MinimallyVerbal Children With Autism: A SequentialMultiple Assignment Randomized Trial

Connie Kasari, PhD, Ann Kaiser, PhD, Kelly Goods, PhD, Jennifer Nietfeld, MA,Pamela Mathy, PhD, Rebecca Landa, PhD, Susan Murphy, PhD, Daniel Almirall, PhD

Objective: This study tested the effect of beginning treatment with a speech-generatingdevice (SGD) in the context of a blended, adaptive treatment design for improvingspontaneous, communicative utterances in school-aged, minimally verbal children withautism. Method: A total of 61 minimally verbal children with autism, aged 5 to 8 years, wererandomized to a blended developmental/behavioral intervention (JASPþEMT) with or withoutthe augmentation of a SGD for 6monthswith a 3-month follow-up. The intervention consisted of 2stages. In stage 1, all children received 2 sessions per week for 3 months. Stage 2 intervention wasadapted (by increased sessions or adding the SGD) based on the child’s early response. The pri-mary outcome was the total number of spontaneous communicative utterances; secondary mea-sures were the total number of novel words and total comments from a natural languagesample. Results: Primary aim results found improvements in spontaneous communicativeutterances, novel words, and comments that all favored the blended behavioral intervention thatbegan by including an SGD (JASPþEMTþSGD) as opposed to spoken words alone(JASPþEMT). Secondary aim results suggest that the adaptive intervention beginning withJASPþEMTþSGD and intensifying JASPþEMTþSGD for children who were slow respondersled to better posttreatment outcomes. Conclusion: Minimally verbal school-aged childrencan make significant and rapid gains in spoken spontaneous language with a novel, blendedintervention that focuses on joint engagement and play skills and incorporates an SGD. Futurestudies should further explore the tailoring design used in this study to better understand children’sresponse to treatment. Clinical trial registration information—Developmental and AugmentedIntervention for Facilitating Expressive Language (CCNIA); http://clinicaltrials.gov/;NCT01013545. J. Am. Acad. Child Adolesc. Psychiatry, 2014;53(6):635–646. Key Words: autismspectrum disorders, minimally verbal, school-aged, communication intervention, SMART design

ommunication impairment is a core deficitin children diagnosed with autism spec-
C trum disorders (ASD). Although the ma-
jority of children learn to communicate withspoken language, approximately 25% to 30% ofchildren with ASD remain minimally verbal,even after years of intervention.1,2 Exact numbersare unknown, largely because research studiesoften exclude children because of limited verbalabilities.1 Failure to develop spoken languageby age 5 years increases the likelihood of a poor

This article is discussed in an editorial by Dr. Helen Tager-Flusberg on page 612.

This article can be used to obtain continuing medical education(CME) at www.jaacap.org.

OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATR

E 53 NUMBER 6 JUNE 2014

long-term prognosis for social and adaptivefunctioning.2,3

Some children can learn spoken language afterthe age of 5 years, but the window of opportunitymay be small.4 A recent review of studies oflanguage acquisition in individuals with ASDreported on 167 individuals who started speakingafter age 5 years.5 The majority of individualswho acquired spoken language did so between5 and 7 years of age and had nonverbal IQs of>50. These individuals often received behavioralinterventions targeting production of sounds andwords and learned to produce single words torequest needs and wants. Only one-third whobegan to use spoken language progressed tophrase speech. Because participant and outcomedescriptors were often limited, the extent to

Y

www.jaacap.org 635

http://clinicaltrials.gov/

http://www.jaacap.org

KASARI et al.

which word production was communicative (i.e.,socially directed to others) is unknown.

One approach to providing minimally verbalchildren a means to communicate is to use au-gmentative or alternative communication (AAC)approaches, most often a picture symbol systemor speech-generating device (SGD). AlthoughAACintervention studies demonstrate improvements incommunication, few have demonstrated changesin spoken language. For example, the Picture Ex-change Communication System (PECS) is a visu-ally based, augmentative communication system inwhich children exchange pictures to communicatewith others. One study randomized 84 children toPECS or control conditions and found that childrenwith PECS training initiated communicative re-quests at a higher rate.6 Vocalizations also im-proved, especially for children who had somespoken language at baseline.7 Language testscores, however, did not improve.6

Another AAC intervention approach involvesan SGD. SGDs display symbols that producevoice output communication when selected. Areview of 23 studies that used an SGD included atotal of 51 children with ASD between the ages of3 and 16 years.8 All studies were single-subjectdesigns, and most focused on teaching, request-ing, or responding to questions using the SGD.Few studies assessed maintenance and general-ization. While using an SGD appears to increasecommunication, particularly requesting in indi-vidual children with ASD,8 no rigorous groupdesigns have replicated these findings, and fewstudies have demonstrated varied communica-tive functions beyond requesting (e.g., com-menting) or an increase in spoken language.Because of the importance of increasing social useof spoken language, in the current study ourprimary outcome measure was total sponta-neous, communicative utterances (SCU) codedfrom a standardized natural language sample(NLS). SCUs are unprompted, generative (non-scripted) communicative utterances that aredirected to a partner for the purpose of sharinginformation (comments), requests, and questions.

Given the lack of spoken language progressfor some children with ASD who have had ac-cess to early intervention services, we consid-ered novel approaches to intervention in thisstudy. We blended 2 communication-focusedand evidence-based early interventions for pr-eschool children—JASPER (Joint Attention Sy-mbolic Play Engagement and Regulation)9,10

and EMT (Enhanced Milieu Teaching)11,12

JOURN

636 www.jaacap.org

hereafter referred to JASPþEMT. JASPER is anaturalistic behavioral intervention focused onthe development of prelinguistic gestures (jointattention, requesting) and play skills within thecontext of play-based interactions as a means toincrease joint engagement between an adultand child with ASD.9,10 EMT is a naturalisticbehavioral intervention that uses responsiveinteraction and systematic modeling andprompting to promote spontaneous, functionalspoken language.11,12 Both JASPER and EMThave shown efficacy in preschool-aged, mini-mally verbal children with ASD.10,12-13

Furthermore, given the promising but limiteddata on the effectiveness of SGDs for childrenwith ASD, we sought to understand the role ofSGDs as a treatment component in the context ofJASPþEMT. Because not all children were ex-pected to benefit equally from these components,we used adaptive intervention designs.14 In anadaptive intervention, treatment may be adapted(e.g., by intensifying the dosage or augmentingthe spoken intervention with SGD) to addressthe specific needs of the child (e.g., if the child ismaking slow progress in spoken communication).

The overarching aim of this study was toconstruct an adaptive intervention that usedJASPþEMT and varied the addition of an SGDwith minimally verbal school-aged children. Threeadaptive interventions were considered in thecontext of a sequential multiple assignment ran-domized trial (SMART), as follows15-19: a firstadaptive intervention that began with JASPþEMTand intensified JASPþEMT for children who wereslow responders; (b) a second adaptive interven-tion that began with JASPþEMT and augmentedJASPþEMT with SGD for children who were slowresponders; and (c) and a third adaptive inter-vention that began with JASPþEMTþSGD andintensified JASPþEMTþSGD for children whowere slow responders. The SMART design ad-dressed 2 aims. The primary aim was to examinethe effect of the adaptive intervention beginningwith JASPþEMTþSGD versus those beginningwith JASPþEMT alone. A secondary aim was tocompare outcomes across the aforementioned3 adaptive interventions.

METHODStudy DesignThis study was a longitudinal (repeated outcomemeasures at baseline and weeks 12, 24, and 36), 3-siteSMART design. This SMART included 2 stages oftreatment (Figure 1). Each stage of treatment was

AL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY

VOLUME 53 NUMBER 6 JUNE 2014


SMART FOR MINIMALLY VERBAL CHILDREN WITH ASD

12 weeks in duration. At the beginning of Stage 1(baseline), all children meeting inclusion criteria wererandomized with equal probability to JASPþEMTversus JASPþEMTþSGD. At the end of 12 weeks,children were assessed for early response versus slowresponse (defined in section “Stage 2 Treatments:Weeks 13 to 24”) to Stage 1 treatment. At the begin-ning of Stage 2 (i.e., beginning of week 13), the sub-sequent treatments were adapted based on responsestatus. All early responders continued with the sametreatment for another 12 weeks. For slow respondersto JASPþEMTþSGD, treatment was intensified(3 sessions per week). Slow responders to JASPþEMTwere re-randomized with equal probability to inten-sified JASPþEMT or augmented JASPþEMTþSGD(Figure 1). The institutional review board at eachsite approved the study protocol. Randomizationwas conducted by an independent data-coordinatingcenter.

FIGURE 1 Participant flow through trial. Note: JASPþEMT ¼JASPþEMTþSGD ¼ spoken mode of JASPER plus Enhanced

JOURNAL OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATR


ParticipantsInclusion criteria were as follows: previous clinicaldiagnosis of ASD, confirmed by research-reliable staffusing the Autism Diagnostic Observational Schedule(ADOS-Generic)20 Module 1 appropriate for childrenwithout phrase speech; chronological age between 5and 8 years; evidence of being minimally verbal, withfewer than 20 spontaneous different words used dur-ing the 20-minute NLS; at least 2 years of prior inter-vention, per parent-report; and receptive language ageof at least 24 months (based on performance of 2 of 3assessments, given the potential difficulty complyingwith standardized test conditions. Exclusion criteriawere as follows: major medical conditions other thanASD; sensory disabilities (e.g., deafness); motor dis-abilities (e.g., cerebral palsy); uncontrolled seizuredisorders; and proficient use of an SGD based onparent-report and observation during study adminis-tration of the NLS.

spoken mode of JASPER plus Enhanced Milieu Teaching;Milieu Teaching plus Speech Generating Device.

Y

www.jaacap.org 637


KASARI et al.

A total of 61 children were randomized in theSMART. At baseline, participants completed diag-nostic and standardized cognitive and language as-sessments. Of the children, 60 met ADOS20 criteria forautism, and 1 met a criterion for ASD. Participantswere a mean age of 6.31 years (SD ¼ 1.16) and had anaverage of 17.23 (SD ¼ 16.44) different words on thebaseline language sample (NLS) administered by anunfamiliar and blinded assessor. Five childrenscreened with more than 20 words (range, 26–51words) and were included because of low intelligi-bility and predominance of scripted language.

TABLE 1 Participant Baseline Demographic and Developmen

TotalN ¼ 61

% n %

GenderMale 83 51 87Female 17 10 13

RaceWhite 48 29 47African-American 23 14 21Asian American 19 12 21Hispanic 5 3 4Other 5 3 7

Maternal education� High school 5 3 3College 38 23 47Graduate school 57 35 50

SiteUCLA 39 24 40VU 33 20 33KKI 28 17 27

Mean SD Me

Age (y) 6.31 1.16 6.Language sample

TSCU 29.44 25.37 28.TDWR 17.20 16.44 16.TCOM 6.02 8.17 7.

Standardized assessmentsTELD-3Receptive AE 2.03 .62 1.Expressive AE 1.73 .39 1.

PPVT-4AE 2.64 .67 2.

Leiter-RBrief IQ 68.18 18.96 68.AE 4.00 1.12 3.

ADOS 20.02 4.37 20.

Note: p Values examine whether significant differences between first-stage treatFisher’s exact test was used. ADOS ¼ Autism Diagnostic Observation Scheduplus Enhanced Milieu Teaching; JASPþEMTþSGD ¼ spoken mode of JASPEKennedy Krieger Institute; PPVT-4 ¼ Peabody Picture Vocabulary Test, 4th EditTELD-3 ¼ Test of Early Language Development, 3rd Edition; TSCU ¼ TotalAngeles; VU ¼ Vanderbilt University.

JOURN

638 www.jaacap.org

Nonverbal cognitive scores averaged 4.00 years (SD ¼1.12), with an average Brief-IQ standard score of 68.18(SD ¼ 18.96). Table 1 shows the baseline (pretreat-ment) characteristics in the overall sample (n ¼ 61)and for stage 1 treatment groups. Baseline childcharacteristics did not differ by treatment groupassignment.

Intervention ProceduresThe core intervention in all components was theJASPþEMT naturalistic communication intervention

tal Variables, by Initial Treatment Assignment

JASPþEMTn ¼ 30

JASPþEMTþSGDn ¼ 31

p Valuen % n

.4426 79 254 21 6

.9214 48 156 25 87 16 51 7 22 4 1

.171 8 3

14 27 815 65 20

.9812 39 1210 32 108 29 9

an SD Mean SD p Value

18 1.08 6.44 1.23 .37

37 29.96 30.48 20.35 .7578 19.52 17.61 12.91 .2701 11.00 5.07 3.80 .86

94 .52 2.12 .71 .2670 .37 1.75 .42 .68

55 .66 2.72 .68 .34

73 21.26 67.65 16.77 .8393 1.12 4.07 1.14 .6360 4.47 19.55 4.27 .35

ment assignments exist. c2 Tests were used, except in cases where n < 5,le; AE ¼ “age equivalent” in years; JASPþEMT ¼ spoken mode of JASPERR plus Enhanced Milieu Teaching plus Speech Generating Device; KKI ¼ion; TCOM ¼ total number comments; TDWR ¼ total different word roots;Socially Communicative Utterances; UCLA ¼ University of California, Los





that taught joint attention, symbolic play, and social useof language during child-preferred play activities. Alltraining occurred in university clinic playrooms. Eachchild was assigned a therapist (speech clinician, specialeducator, or child psychologist) who was trained tocriterion fidelity (>0.90) on all elements of the treat-ment variations. During the first phase of the inter-vention (12 weeks), all children received 24 sessions,1 hour each, of treatment in either the JASP þ EMTspoken mode or JASPþ EMTþ SGD mode. During theJASPþEMTþSGD condition, the SGD was used tomodel a minimum of 50% of all spoken communica-tion. In the second phase of the intervention, parentswere included in the second 24 treatment sessions.Following a manualized protocol, the child’s therapistimplemented systematic parent training consistentwith the treatment variation to which the child wasassigned. Throughout both treatment phases and alladaptive treatment variations, fidelity of therapistimplementation was assessed for 20% of the sessions.

MeasuresNatural Language Sample. The NLS was a 20-minutestandardized, naturalistic interaction in which anadult and child played with a specific set of toys. Theadult was responsive to child verbal and nonverbalcommunication but did not prompt the child to talk.The NLS provides a standard context (time, materials,interaction style) that can be used to evaluate a child’sspontaneous expressive language ability.21 Standard-ized NLS can be used to collect repeated measures ofchild language production to index growth over time.Such measures provide stable estimates of childexpressive abilities, are appropriate for repeated mea-sures, and have been shown to be sensitive to changesassociated with language interventions.21 NLS-basedmeasures have been recommended to index produc-tive language abilities of children with autism relativeto age-typical normative development.21,22

Research staff blind to treatment condition admin-istered the NLS; staff were trained to 90% fidelity cri-terion on the NLS interaction procedures prior toassessment. A 24-item checklist based on proceduresused in prior research studies23 was used to assess fi-delity. Fidelity was rated on 20% of the sessions byindependent and blinded coders. Staff transcribed thesamples using Systematic Analysis of Language Tran-scripts22 (SALT) conventions. Separate, blinded codersverified the transcripts and coded each child utterancefor generativity (not scripted), and communicativefunction (e.g., comment, request, other). Both spokenand SGD-produced utterances were transcribed andcoded, and mode was noted. Reliability of coding ofutterances was determined using an exact agreementformula for each individual code that was thenaggregated across codes; overall reliability was 88.1%.

From the NLS, the primary outcome was the totalnumber of spontaneous communicative utterances



(TSCU), which included comments, requests, and pr-otests and excluded scripted and nonsocial utterances.Secondary outcomes were total number of differentword roots (TDWR) and number of comments (TCOM).Both spoken and SGD-produced utterances were in-cluded for all variables. Measures were derived fromNLS collected at baseline (pretreatment) and at weeks12, 24, and 36 (follow-up). Along with TSCU, TDWR,and TCOM, data on the following additional languagevariables were collected to derive the early/slow-response measure used to trigger stage 2 treatments:proportion of all utterances that were socially commu-nicative (PSCU), words per minute (WPM), meanlength utterance in words (MLUw), and number ofunique word combinations (NUWC).

Intervention Session Transcripts. Intervention sessionswere transcribed and coded following the same con-ventions and including the same 7 variables (TSCU,TDWR, TCOM, PSCU, WPM, MLUw, and NUWC) aswith the NLS. Ten-minute sections (minutes 2–12) weretranscribed during the first week of intervention andthe 12th week of intervention. These 7 session transcriptmeasures were used (along with the 7 from the NLS)for determining early/slow response to treatment.

Autism Diagnostic Observation Schedule, Module 1. TheADOS is a 30- to 45-minute, semistructured, play-basedassessment from which operationally defined behaviorsassociated with ASD are rated on a scale of 0 to 2 or 3points. An algorithm score provides cut-off scores forASD and autism classifications. All participantscompleted the ADOS Module 1 of the ADOS, designedfor children at the nonverbal or single-word phase. TheADOS was completed at baseline to confirm eligibility.

Leiter International Performance Scale–Revised. TheLeiter International Performance Scale–Revised (Leiter-R)is a nonverbal cognitive assessment for individuals 2 to20 years of age. Tasks include matching, pattern comple-tion, and sequential order, and do not require verbal re-sponses. This test, completed at baseline to confirmeligibility, yields global IQ and mental age equivalencescores.

Peabody Picture Vocabulary Test, Fourth Edition. ThePeabody Picture Vocabulary Test, Fourth Edition(PPVT-4) is a test of receptive vocabulary develop-ment that is appropriate for children aged 2 yearsand older. From an array of 4 pictures, the childidentifies (by pointing) the picture that best illus-trates the word pronounced by the examiner. ThePPVT-4 was completed at baseline to determineeligibility, and yields age-equivalent and standardscores.

Test of Early Language Development, Third Edition. TheTest of Early Language Development, Third Edition(TELD-3) is a standardized assessment with subtests forreceptive and expressive vocabulary development. Theassessment is normed for children 2 or more yearsof age. The TELD-3, given at baseline, yields age-equivalent and standardized scores; the receptive sub-test was used to confirm eligibility.

Y

www.jaacap.org 639


KASARI et al.

Demographic Questionnaire. Parents completed abrief demographic questionnaire providing infor-mation about the child and family, includingchild’s previous early intervention and currentservices.

Intervention StagesThere were 2 intervention stages (early and adapted)in the study, for a total of 24 weeks of intervention.Interventionists were trained to criterion across sitesand were closely supervised both on-site and throughsite visits, weekly conference calls, video feedback,and fidelity checks for 20% of sessions that were ratedusing an exact agreement formula (agreements di-vided by agreements plus disagreements). Fidelity oftreatment implementation averaged 94.26% (SD ¼5%) for JASPþEMT and 93.69% (SD ¼ 4%) forJASPþEMTþSGD.

Stage 1 Treatments: Weeks 1 to 12JASPþEMT. Participants initially randomized to thiscondition received two 1-hour sessions per week, for12 weeks. JASP (based on JASPER) focuses on earlysocial-communication skills, including coordinatedjoint attention gestures known to be associated withthe development of later spoken language of childrenwith autism.9,10 Intervention ingredients include thecreation of contextually relevant and meaningfullearning opportunities during interactions with adultpartners (therapists, parents) who are responsive tochild interests and actions, who model and expandplay and gesture use and maintain joint engagement.

FIGURE 2 Primary aim results for the primary outcome (totcharacters denote observed means; closed denote model-estimestimated means. JASPþEMT ¼ spoken mode of JASPER plusmode of JASPER plus Enhanced Milieu Teaching plus Speech

JOURN

640 www.jaacap.org

The second intervention focuses on spoken languageacquisition, EnhancedMilieu Teaching (EMT).11,12 EMTis a naturalistic early language intervention that uses 7core strategies to teach language in social interaction:following the child’s lead in conversation and play,responding to communicative initiations from the childwith target language, expanding child utterances byadding words to increase complexity while maintainingthe child’s meaning, arranging the environment tosupport and elicit communication from the child, andsystematic use of prompts (model, time delays, andprompts). Both JASPER and EMT are manualized; theblended intervention was called JASPþEMT.

JASPþEMTþSGD. Participants initially random-ized to this condition received 2 sessions per week, each1 hour in length, for 12 weeks. All aspects of the inter-vention were the same as for JASPþEMT, except theaddition of an SGD such as an iPad or DynaVox. TheSGD was programmed with vocabulary relevant to thetoys or activities used during each treatment session.The interventionist modeled and expanded target lan-guage on the SGD in conjunction with spoken languagefollowing the modeling and expansion protocols ofJASPþEMT; the SGD intervention followed a writtenprotocol. To ensure fidelity of use of the SGD, theinterventionist was required to use the SGD at least 50%of the time when modeling language. If the child initi-ated a communication bid using the SGD, the inter-ventionist expanded this bid at least 80% of the time.

Stage 2 Treatments: Weeks 13 to 24Early/Slow-Response Measure. Stage 2 treatmentsdepended on early versus slow response to Stage 1

al social communicative utterances). Note: Open plottingated means. Error bars denote 95% CI for model-Enhanced Milieu Teaching; JASPþEMTþSGD ¼ spokenGenerating Device.





treatments (JASPþEMT or JASPþEMTþSGD). Earlyversus slow response was defined based on 14 mea-sures from 2 sources: the 7 communication variables(TSCU, PSCU, MLUw, TDWR, WPM, TCOM, andNUWC) from the NLS with blinded assessor, and thesame 7 communication variables from the interven-tion transcripts. For each of the 14 variables, wecalculated the percent change from baseline to week12. If the child demonstrated 25% or greater changeon at least one-half of the variables (i.e., 7 of 14), thenthe participant was considered an early responder;otherwise, the child was considered a slowresponder. Early responders continued their stage 1intervention assignment for an additional 12 weeks instage 2.

Children who were slow responders at the end ofstage 1 received a modified intervention to improveoutcomes in spontaneous communication. Slow re-sponders to stage 1 JASPþEMT intervention werere-randomized with equal probability to either in-tensified JASPþEMT (increased dose of intervention)or the addition of the SGD (augmenting the inter-vention). Slow responders to JASPþEMTþSGD in-tervention for Stage 1 were assigned to intensifiedJASPþEMTþSGD.

Intensified JASPþEMTþSGD. This intervention wasidentical in content to JASPþEMTþSGD but occurredfor a total of 3 hours per week for an additional 12weeks.

Intensified JASPþEMT. This intervention was iden-tical in content to JASPþEMT but occurred for a total of3 hours per week for another 12 weeks.

Augmented JASPþEMT. This intervention wasidentical to the Stage 1 JASPþEMTþSGD intervention.It consisted of two 1-hour sessions per week foranother 12 weeks.

Parent Participation in the InterventionParents were involved in all stages of the study. Duringstage 1, parents watched intervention sessions through1-way mirrors. In Stage 2, all parents were providedtraining concurrent with the intervention sessions in all3 adapted conditions. Parents joined the child andtherapist in the clinic room, observed the therapistworking with the child for part of each session, andthen practiced the intervention with the child while thetherapist coached the parents’ implementation. Fidelityof parent implementation was assessed in 20% of thesessions and averaged 67.38 (SD ¼ 11.07) for parents inJASPþEMT and 66.46 (SD ¼ 13.08) for parents inJASPþEMTþSGD.

Study AimsThe primary aim of study was to examine the effec-tiveness of the adaptive intervention beginning withJASPþEMTþSGD versus beginning with JASPþEMTon longitudinal outcomes at weeks 12, 24 (primaryendpoint), and 36 (follow-up). This corresponds to



testing the main effect of stage 1 treatment (initialJASPþEMT versus initial JASPþEMTþSGD). The sec-ondary aim of study was to compare mean outcomes atweeks 24 and 36 among the 3 embedded adaptiveinterventions.

Sample SizeThe planned sample size was based on the pri-mary aim, using the primary outcome (TSCU), abetween-groups comparison of stage 1 treatment(JASPþEMTþSGD versus JASPþEMT) of the av-erage TSCU at week 24 (the primary endpoint).Using a 2-sided, 2-sample t test with a type I errorrate of 5%, and assuming an attrition rate of 10%by week 24, the planned total sample size for thisstudy was n ¼ 97 to detect a moderate effect sizeof 0.6 in TSCU with at least 80% power.

Statistical AnalysisLongitudinal regression models were used to ex-amine mean differences in the primary outcome(TSCU) and secondary outcomes (TDWR, TCOM)between the 2 stage 1 treatments (JASPþEMTþSGDversus JASPþEMT) at weeks 0, 12, 24, and 36.Separate models were fit for each longitudinal out-come. Model diagnostics were used to choose modelspecifications for time (in weeks): for TSCU, aquadratic model for JASPþEMTþSGD and a linearmodel for JASPþEMT fit best (Figure 2); for TDWRand TCOM, piecewise linear models with knot atweek 12 fit best. Each model included the followingbaseline covariates: age (years), gender (female asreferent), ethnicity (indicators for African American,white, Hispanic, Asian, with other as the referent),site (indicators for sites 1 and 2, site 3 as referent),and total ADOS at baseline. In each model, the re-sidual error terms were assumed to follow a mean-0 normal distribution with a compound symmetriccovariance structure used to capture the within-person correlation over time (except for TCOM, forwhich an unstructured covariance structure led tobetter fit). Fitted models were used to calculate (andto plot) mean scores (marginal over baseline cova-riates) at each time point, and to report between-groups comparisons at weeks 12 and 24 (primaryendpoint) and 36 (follow-up). We also report within-treatment group change from baseline to week 24(during treatment) and from week 24 to week 36(posttreatment follow-up). In post hoc analyses, weexamined site � time � treatment interaction effects.

For the secondary aim analysis, a weighted re-gression16 was used to compare means in the pri-mary and secondary outcomes among the 3embedded adaptive interventions at weeks 24 and36. A separate model was fit for each outcome. Eachmodel included an indicator for time (week 24as referent), an indicator for stage 1 treatment(JASPþEMT versus JASPþEMTþSGD), an indicator

Y

www.jaacap.org 641


TABLE 2 Estimated Means for Primary Total Socially Communicative Utterances (TSCU) and Secondary Total DifferentWord Roots (TDWR) and Total Number Comments (TCOM) Outcomes at Weeks 12, 24, and 36, by Stage 1 TreatmentAssignment

JASPþEMT JASPþEMTþSGD Difference

Mean 95% CI Mean 95% CI Mean 95% CI Effect Size

Week 12TSCU 35.26 (28.58, 41.94) 54.40 (45.94, 62.86) 19.14* (10.85, 27.44) 0.57TDWR 24.32 (16.87, 31.78) 33.11 (26.94, 39.27) 8.78* (.44, 17.12) 0.34TCOM 8.10 (4.83, 11.37) 14.09 (10.88, 17.30) 5.99* (1.62, 10.36) 0.51

Week 24TSCU 40.34 (32.67, 48.01) 61.90 (52.80, 71.00) 21.56* (10.70, 32.42) 0.62TDWR 25.62 (19.11, 32.14) 33.11 (26.94, 39.27) 7.48* (.20, 14.77) 0.29TCOM 8.10 (4.83, 11.37) 14.09 (10.88, 17.30) 5.99* (1.62, 10.36) 0.44

Week 36TSCU 45.42 (65.64, 55.19) 52.68 (42.58, 62.77) 7.26 (e6.16, 20.68) 0.22TDWR 26.93 (18.62, 35.23) 33.11 (26.94, 39.27) 6.18* (2.44, 14.40) 0.21TCOM 8.10 (4.93, 11.37) 14.09 (10.88, 17.30) 5.99* (1.62, 10.36) 0.54

Note: JASPþEMT ¼ spoken mode of JASPER plus Enhanced Milieu Teaching; JASPþEMTþSGD ¼ spoken mode of JASPER plus Enhanced Milieu Teachingplus Speech Generating Device.*p < .05.

KASARI et al.

for stage 2 treatment (intensify JASPþEMT versusaugment with JASPþEMTþSGD) nested withinJASPþEMT, and time � treatment interaction terms.Each model included the baseline covariates notedabove, plus baseline TSCU, PSCU, TDWR, andTCOM. Nonresponders to JASPþEMT were assigneda weight of 4 to account for having 1 in 4 chances(nonresponders to JASPþEMT were randomizedtwice with probability of 1 in 2 chances) of followingtheir assigned sequence of treatments. All otherchildren were assigned a weight of 2. Robust stan-dard errors, which account for sampling variation inthe distribution of the weights, were used. We alsoreport the rate of early or slow response at week 12by stage 1 treatment assignment.

All randomized participants were included in allanalyses in accordance with intention-to-treat princi-ples. To enhance interpretation of the results for theprimary and secondary aims, we report treatment ef-fect sizes (Cohen’s d)24 at each time point, defined asthe estimated mean difference divided by the standarddeviation in the outcome. For the response rate at week12, we report the number of children who needed to betreated (NNT)25 initially with JASPþEMTþSGD ratherthan JASPþEMT for one additional child to benefit. Avalue of p < .05 (2-sided) was considered statisticallysignificant; 95% confidence intervals (95% CI) werecalculated for all point estimates. Data were analyzedusing the nlme (primary aim) and the geepack (second-ary aim) packages in R.

Missing DataMultiple imputation was used to replace missing valuesin the outcomes and other measures. A sequentialregression multivariate imputation algorithm was

JOURN

642 www.jaacap.org

implemented using the mice package for R. The impu-tation model used was congenial with all analysismodels: it included all longitudinal outcome measures,treatment indicators, early/slow-response measures atweek 12, the baseline covariates listed above, and othertime-varying measures thought to be correlated withoutcomes. Twenty imputed data sets were generated.Point estimates, standard errors, and all tests werecalculated using Rubin’s rules for combining the re-sults of identical analyses performed on each of the 20imputed data sets. Sensitivity analyses26 were con-ducted to assess the robustness of the results to themissing-at-random assumption. This assumptionstates that, given the observed data included in theimputationmodel, the reason for missing data does notdepend on data that are not observed.

RESULTSRecruitment and RetentionOf the 134 children examined for eligibility, 61children met criteria and were randomized atstage 1 (Figure 1). Because of difficulties inrecruiting participants who met the inclusioncriteria, it was not possible to meet the totalplanned sample size of 97 (see Discussion). Manyof the ineligible participants were below thedevelopmental cut-off of 24 months. All missingvalues in this study were due to participantattrition from the study; attrition rates were 10%by week 12, 14% by week 24, and 25% by week36. Attrition did not differ by stage 1 randomizedtreatment assignment (p ¼ .71). Among partici-pants still in the study at week 12, attrition




TABLE3

Estim

ated

Mea

nsforTotalS

ocially

Com

mun

icativeUtterances

(TSC

U,Prim

aryOutcome),TotalD

ifferen

tWordRo

ots(TDW

R)an

dTotalN

umbe

rof

Com

men

ts(TCOM)at

Weeks

24an

d36,forEa

chof

theThreeAda

ptiveInterven

tions

(AI)

AI1

aAI2

AI3

Differen

ces

AI3

vsAI1

AI3

vsAI2

Mea

n95

%CI

Mea

n95

%CI

Mea

n95

%CI

Mea

n95

%CI

Effect

Size

Mea

n95

%CI

Effect

Size

Wee

k24

TSCU

42.74

(33.22

,52.27

)39

.58

(28.50

,50.66

)58

.45

(46.71

,70.18

)15

.70

(e0.45

,31.85

)0.45

18.87*

(1.45,

36.29)

0.55

TDW

R27

.51

(21.65

,33.38

)26

.46

(19.48

,33.43

)33

.48

(26.05

,40.91

)5.97

(e3.53

,15.47

)0.23

7.03

(e3.78

,17

.83)

0.27

TCOM

10.79

(7.44,

14.14)

9.25

(5.85,

12.66)

11.90

(8.10,

15.70)

1.11

(e4.05

,6.27

)0.08

2.65

(e2.70

,8.00

)0.19

Wee

k36

TSCU

45.58

(35.93

,55.24

)44

.51

(32.06

,56.96

)52

.42

(44.37

,60.46

)6.83

(e5.87

,19.53

)0.20

7.91

(7.42e

23.23)

0.23

TDW

R29

.75

(22.85

,36.64

)26

.36

(18.13

,34.59

)31

.94

(26.66

,38.21

)2.19

(e6.53

,10.91

)0.09

5.58

(e4.72

,15

.88)

0.22

TCOM

8.03

(4.70,

11.37)

8.24

(5.25,

11.23)

13.77

(10.21

,17.33

)5.74

*(0.82,

10.67)

0.45

5.53

*(0.80,

10.27)

0.44

Note:

JASP

þEMT¼

spoken

mod

eof

JASP

ERplus

Enha

nced

Milieu

Teac

hing

;JASP

þEMTþ

SGD¼

spoken

mod

eof

JASP

ERplus

Enha

nced

Milieu

Teac

hing

plus

Spee

chGeneratingDevice.

aThead

aptiveinterventions

(AI)arede

fined

asfollows:

AI1

begins

treatmentw

ithJASP

þEMT,

then

intensifies

JASP

þEMTforslo

wrespon

ders

(AþB

inFigu

re1);AI2

begins

treatmentw

ithJASP

þEMT,

then

augm

ents

JASP

þEMTwith

SGDforslo

wrespon

ders(AþC

);AI3

begins

treatmentw

ithJASP

þEMTþ

SGD,then

intensifies

JASP

þEMTþ

SGDforslo

wrespon

ders(DþE

).Allvaria

bles

arefre

quency

counts.

*p<

.05.




during the follow-up period (weeks 24–36) didnot differ by early/slow-response status (p ¼ .86)or by stage 2 randomized treatment assignmentamong slow responders to JASPþEMT (p ¼ .89).In sensitivity analyses26 concerning the missingdata, results reported below were robust to vio-lations of the missing-at-random assumption.

Primary Aim: Main Effect of Stage 1 Treatment(JASPþEMTþSGD Versus JASPþEMT)Primary Contrast. Intervening with JASPþEMTþSGD initially (versus starting with JASPþEMTalone) led to greater TSCU at week 24 (p < .01)(Table 2). Specifically, the average TSCU at week24 for JASPþEMTþSGD was 61.9 utterances(95% CI ¼ 52.8–71.0) versus 40.3 utterances (95%CI ¼ 32.7–48.0) for JASPþEMT, a clinically sig-nificant average difference of 21.6 utterances (95%CI ¼ 10.7–32.4). This difference is approximatelydouble the rate of communicative utterance perminute from baseline, which is considerable forthis sample of low-rate communicators. Thiscorresponds to a moderate-to-large treatmenteffect size of 0.62. On average, 92.1% (95% CI ¼87.8%–96.5%) of the total TSCU (across all timepoints) in the JASPþEMTþSGD group werespoken utterances. In additional analyses inwhichonly spoken utterances were used for the JASPþEMTþSGD group, the effect of SGD on TSCU atweek 24 was attenuated (effect size of 0.51) butremained statistically significant (p < .05).

Additional Contrasts. Similarly, for the second-ary outcomes, Stage 1 JASPþEMTþSGD led togreater TDWR (p ¼ .04) and TCOM (p < .01) atweek 24. Treatment effect sizes for the secondaryoutcomes at week 24 were small (0.29 for TDWR)to small-to-moderate (0.44 for TCOM). For alloutcomes, stage 1 JASPþEMTþSGD was supe-rior to JASPþEMT at week 12. For TDWR andTCOM, treatment effects peaked at week 12,whereas for TSCU, effects peaked at week 24. Byweek 36, treatment effects were maintained forTCOM, but attenuated for TSCU and TDWR. Inpost hoc analyses, there was no evidence of sig-nificant site � treatment interaction effects at anytime point (TSCU, p ¼ .45; TDWR, p ¼ 0.78;TCOM, p ¼ .28).

Secondary Aim: Comparison of EmbeddedAdaptive InterventionsConsistent with the results for the primary aim,the adaptive interventions leading to the great-est TSCU at week 24 were the adaptive in-terventions that began with JASPþEMTþSGD

Y

www.jaacap.org 643


KASARI et al.

and intensified JASPþEMTþSGD among chil-dren who were slow responders (Table 3).Among the 2 adaptive interventions beginningwith JASPþEMT, the adaptive intervention thataugmented JASPþEMT with SGD among slowresponders led to greater TSCU (42.7; 95% CI ¼33.2–52.3) than the adaptive intervention, whichintensified JASPþEMT for slow responders(39.6; 95% CI ¼ 28.5–50.7); however, the meandifference in outcome was not significant eitherclinically or statistically (ES, 0.10; 95% CI ¼ �4.2to 10.5; p ¼ .40). Results were similar at week 36.

The overall rate of early response for all par-ticipants at week 12 was 70% (95% CI ¼ 57.9%–

82.3%). Children assigned to JASPþEMT had aresponse rate of 62.2% (95% CI ¼ 45%–79.3%),whereas children assigned to JASPþEMTþSGDhad a response rate of 77.7% (95% CI ¼ 60.6%–

95.0%). This 15.6% difference is clinically, but notstatistically, significant (95% CI ¼ �8.7% to39.9%, p ¼ 0.20). Approximately 6.5 childrenneeded to be treated26 (95% CI, number neededto harm [NNTH] 11.5 to N to number neededto treat for benefit [NNTB] 2.5) initially withJASPþEMTþSGD rather than JASPþEMT alonefor one additional child to respond by week 12.

DISCUSSIONThe current study focused on increasing sponta-neous communicative, spoken language in mini-mally verbal, school-aged children with ASD.Using a novel blended, adaptive intervention,children improved over a 6-month treatment and3-month follow-up. These findings are particularlyimportant because the intervention was providedto children who were minimally verbal after earlyintervention, and, in most cases, after at least 2years of early intensive behavioral interventions.Children showed significant gains in spontaneouscommunication in a short period of time in a rela-tively low-intensity developmental and behavior-ally based intervention of 2 to 3 hours per week.

There were 3 main findings.First, there was a robust and consistent finding

that beginning intervention with the SGD inte-grated into the blended intervention was superiorin producing more spontaneous communicativeutterances than beginning intervention with theblended intervention and spoken language only.These data are especially important given ourcurrent knowledge of effective interventions forminimally verbal children with ASD. One ran-domized trial and several single case studies have

JOURN

644 www.jaacap.org

found benefit in augmentative communicationapproaches for increasing requesting, but notelimitations on improving other functions ofspoken language. A recent study with preschool-aged, minimally verbal children found gains inspoken language for 2 oral language–based in-terventions.27 This small-scale RCT with 17 mini-mally verbal preschoolers with ASD and withmental ages of more than 12 months did not yieldsignificant effects of 2 behavioral interventions (1naturalistic and 1 discrete trial teaching); rather,both interventions doubled spoken communica-tion during a 20-minute behavior sample, from 2words at baseline to 4 to 5 words at exit. Amoderator analysis found that children who hadmore joint attention skills before treatmentresponded better in both treatments. Joint atten-tion skills have predicted spoken language out-comes in several previous intervention studies,27-29

including 1 study that followed children over a5-year period after receiving JASPER during pre-school.30 Thus, the current study,which focused ondevelopmental prerequisites to spoken languageincluding joint attention, joint engagement, andplay along with systematic modeling and pro-mpting for spoken language, may have providedthe combination of supports needed for minimallyverbal children with ASD to successfully increasetheir spoken communication.

Second, this is 1 of the first studies to showincreases in minimally verbal children’s spon-taneous communication including differenttypes of words and functions beyond request-ing. Positive outcomes in previous studiesgenerally have been limited to increases inrequesting behavior when an augmentativemeans of communication is introduced.7-9 Thefocus on requesting is expected in an adult-directed intervention approach in which thechild is often prompted to comply with in-structions and there are strong external rein-forcement strategies in place. The currentintervention approach used a developmental,child-directed approach with strong naturalisticreinforcement strategies in place. Adults in theintervention were contingently responsive tochild attempts at communication and providedexpansion of language through models thatmatched the child’s communicative intent.Children in the current study demonstrated in-creases in the spontaneous use of language suchas commenting and novel words; these out-comes were significantly greater for the groupbeginning with JASPþEMTþSGD.





Finally, this study a used unique interventiondesign that tailored stage 2 treatment dependenton the child’s response to stage 1 treatment.Adding in the SGD later for slow responders toJASPþEMT alone did not provide the samebenefit as adding in the SGD from the beginning oftreatment. These data are provocative, and sug-gest that including an SGD along with naturalisticbehavioral interventions at the start of treatmentmay be most beneficial to minimally verbal chil-dren. The impact of the SGD on spoken languageoutcomes is consistent with findings from bothsingle case and group design studies.31 Modestincreases in spoken language have been associatedwith SGD and manual sign interventions; how-ever, no previous studies have implemented SGDtraining using naturalistic approaches such as theblended JASPþEMT intervention, and moststudies have been limited to exploring requestingrather than use of spoken language for the fullrange of pragmatic functions. Based on previousresearch,32 3 pathways have been suggested bywhich SGD intervention might promote spokenlanguage: namely, by increasing the frequency ofcommunication; by reducing the motor responsedemands and pressure to communicate; andby altering the acoustic effects of the child’scommunication. Of these potential pathways, weposit that changes in the acoustic signal providedto the child by increasing the number andphonological consistency of models when usingthe SGD and pairing the acoustic signal with thegraphic SGD symbol are likely to have influencedthe outcomes in the current study. Further re-search examining the effects of increased modelsand auditory/visual pairing on spoken languageclearly is needed.

A limitation of the current study is that althoughthe sample size is relatively large for a randomizedtrial in autism research (particularly for this sub-population of children), we enrolled approxi-mately two-thirds of our recruitment target. One ofthe implications of this may have been a limitedability to distinguish between the 2 adaptive in-terventions beginningwith JASPþEMT (secondaryaim). Thus, the fine-grained analysis comparingthe 2 adaptive interventions that differ by stage 2treatment requires replicationwith a larger sample.

Despite this limitation, this is one of the firststudies to develop and evaluate the components ofan adaptive intervention based on children’s earlyresponse for improving spoken language out-comes in the understudied and underserved pop-ulation of minimally verbal school-aged children



with ASD. The results of the study suggest thatimprovements in spontaneous, communicativeutterances, novel words, and comments all favoredthe blended behavioral intervention that beganwith the addition of a SGD (JASPþEMTþSGD) asopposed to JASPþEMT with spoken words only.Secondary aim results suggest that the adaptiveintervention beginning with JASPþEMTþSGDand intensifying the JASPþEMTþSGD treatmentfor children who were slow responders led tobetter posttreatment outcomes. There was insuffi-cient evidence that the adaptive intervention thatintroduces SGD among children who respondslowly to JASPþEMT differed from the adaptiveintervention that intensifies JASPþEMT for thesechildren. It is important that in this sample, chil-dren who had already had an average of 2 years ofprior treatment made progress in spoken languageacross all conditions. There is much more to belearned about effective communication interven-tion for this population. Future studies should testadaptive interventions based on various in-terventions, in an effort to further understandwhatprogress in spoken communication is possible inchildren with autism. &

Y

Accepted March 6, 2014.

Dr. Kasari is with the University of California-Los Angeles (UCLA) SemelInstitute for Neuroscience and Human Behavior. Dr. Kaiser and Ms.Nietfeld are with Vanderbilt University. Dr. Goods is with First Five LosAngeles. Dr. Mathy is with Speech, Language and Hearing Clinic,University of Utah. Dr. Landa is with Kennedy Krieger Institute. Drs.Murphy and Almirall are with University of Michigan.

This study was funded by Autism Speaks #5666, CharacterizingCognition in Nonverbal Individuals with Autism, an initiative begun byMs. Portia Iverson and Cure Autism Now. Grant support was alsoprovided by the National Institute of Child Health and Human Devel-opment (NICHD) R01HD073975-02 (C.K., A.K., S.M., D.A.), andR01HD073975-03 (C.K., R.L.), R03MH097954-02 (D.A.), andRC4MH092722-01 (D.A.) from the National Institute of MentalHealth (NIMH), and P50DA10075 (National Institute on Drug Abuse[NIDA], S.M., D.A.).

Drs.Murphy and Almirall served as the statistical experts for this research.

The authors thank the families and children who participated in this studyand the teamof interventionists, coders, transcribers, anddataanalysts atour 3 sites: Ya-Chih Chang, PhD, Stephanie Patterson, PhD, KathryneKrueger, PhD, Charlotte Mucchetti, PhD, Dalia Kabab, BA, CaitlinMcCracken, BA, Julia Kim, BA, Alison Holbrook, BA, Abbey Hye, BA,and Kelsey Johnson BA, BS, from UCLA; Stephanie Jordan, MA, Court-ney Wright, PhD, Blair Burnette, MA, and Ann Simonson MA, fromVanderbilt University; Philip Menard, MS, Emily Watkins, MS, KerryBuechler, MS, Christine Hess, PhD, Sarah Gardner , BS, and MelissaSubbock, MS from Kennedy Krieger Institute; and Xi Lu, MA from Uni-versity of Michigan.

Disclosure: Dr. Kasari has received salary support from grants from theHealth Resources and Services Administration (HRSA), the National In-stitutes of Health (NIH), and Autism Speaks. Dr. Kaiser has received sup-port from NIH and the Institute of Education Sciences (IES). Dr. Landa hasreceived support from NIMH, the Centers for Disease Control and Pre-vention (CDC), Autism Speaks, the Simons Foundation Autism ResearchInitiative (SFARI), IES,HRSA, and RideOn For Autism (ROAR). Drs.MurphyandAlmirall have received support fromNIH andNIDA.Ms.Nietfeld has

www.jaacap.org 645


64

KASARI et al.

received support from NIH and IES. Drs. Goods and Mathy report nobiomedical financial interests or potential conflicts of interest.

Correspondence to Connie Kasari, PhD, 68-268 Semel Institute, UCLA,Los Angeles, CA 90024; e-mail: [email protected]

JOURNAL

6 www.jaacap.org

0890-8567/$36.00/ª2014 American Academy of Child andAdolescent Psychiatry

http://dx.doi.org/10.1016/j.jaac.2014.01.019

REFERENCES
1. Tager-Flusberg H, Kasari C. Minimally verbal school-aged chil-
dren with autism: the neglected end of the spectrum. Autism Res.2013;6:468-478.

2. Anderson DK, Oti RS, Lord C, Welch K. Patterns of growth inadaptive social abilities among children with autism spectrumdisorders. J Abnorm Child Psychol. 2007;37:1019-1034.

3. Rutter M, Greenfeld D, Lockyer L. A five to fifteen year follow-upof infantile psychosis: II. Social and behavioral outcome. Br JPsychiatry. 1967;113:1183-1189.

4. Wodka E, Mathey P, Kalb L. Predictors of phrase and fluentspeech in children with autism and severe language delay. Pedi-atrics. 2013;131:1-7.

5. Pickett E, Pullara O, O’Grady J, Gordon B. Speech acquisition inolder nonverbal individuals with autism: a review of features,methods and prognosis. Cogn Behav Neurol. 2009;22:1-21.

6. Howlin P, Gordon RK, Pasco G, Wade A, Charman T. The effec-tiveness of picture exchange communication system (PECS)training for teachers of children with autism: a pragmatic grouprandomized controlled trial. J Child Psychol Psychiatry. 2007;48:473-481.

7. Gordon K, Pasco G, McElduff F, Wade A, Howlin P, Charman T.A communication-based intervention for nonverbal children withautism: what changes? Who benefits? J Consult Clin Psychol. 2011;79:447-457.

8. Van Der Meer L, Rispoli M. Communication interventionsinvolving speech-generating devices for children with autism: areview of the literature. Dev Neurorehabil. 2010;13:294-306.

9. Kasari C, Gulsrud AC, Wong C, Kwon S, Locke J. A randomizedcontrolled caregiver mediated joint engagement intervention fortoddlers with autism. J Autism Dev Disord. 2010;40:1045-1056.

10. Kasari C, Paparella T, Freeman S, Jahromi LB. Language outcomein autism: randomized comparison of joint attention and playinterventions. J Consul Clin Psychol. 2008;76:125-137.

11. Kaiser AP, Goetz L. Enhancing communication with personslabeled severely disabled. J Assoc Persons Severe Handicaps. 1993;18:137-142.

12. Kaiser AP, Hancock TB, Nietfeld JP. The effects of parent-implemented enhanced milieu teaching on the social communi-cation of children who have autism. J Early Ed Dev. 2000;11:423-446.

13. Goods K, Ishijima E, Chang Y-C, Kasari C. Preschool basedJASPER intervention in minimally verbal children with autism:pilot RCT. J Autism Dev Disord. 2013;43:1050-1056.

14. Collins LM, Murphy SA, Bierman KA. A conceptual frame-work for adaptive preventive interventions. Prev Sci. 2004;5:185-196.

15. Murphy SA. An experimental design for the development ofadaptive treatment strategies. Stat Med. 2005;24:1455-1481.

16. Collins LM, Murphy SA, Strecher V. The Multiphase OptimizationStrategy (MOST) and the Sequential Multiple Assignment Ran-domized Trial (SMART): new methods for more potent e-healthinterventions. Am J Prev Med. 2007;32:S112-S118.

17. Nahum-Shani I, Qian M, Almirall D, et al. Experimental designand primary data analysis methods for comparing adaptive in-terventions. Psychol Methods. 2012;17:457-477.

18. Almirall D, Compton SN, Rynn MA, Walkup JT, Murphy SA.SMARTer discontinuation trial designs for developing an adaptivetreatment strategy. J Child Adolesc Psychopharmacol. 2012;22:364-374.

19. Almirall D, Compton SN, Gunlicks-Stoessel M, Duan N,Murphy SA. Designing a pilot sequential multiple assignmentrandomized trial for developing an adaptive treatment strategy.Stat Med. 2012;31:1887-1902.

20. Lord C, Rutter M, DiLavore PC, Risi S. Autism DiagnosticObservation Schedule. Los Angeles, CA: Western PsychologicalServices; 1999.

21. Tager-Flusberg H, Rogers S, Cooper J, et al. Defining spoken lan-guage benchmarks and selecting measures of expressive languagedevelopment for young children with autism spectrum disorders.J Speech Lang Hear Res. 2009;52:643-652.

22. Miller J, Iglesias A. SALT: Systematic Analysis of LanguageTranscripts. Software for the analysis of oral language. Middleton,WI: SALT Software; 2012.

23. Kaiser AP, Roberts MY. Parent-implemented enhanced milieuteaching with preschool children with intellectual disabilities.J Speech Lang Hear Res. 2013;56:295-309.

24. Cohen J. Statistical Power Analysis for the Behavioral Sciences. 2nded. Hillsdale, NJ: Lawrence Erlbaum Associates; 1988. xxi, 567.

25. Altman D. Confidence intervals for the number needed to treat. BrMed J. 1998;317:1309-1312.

26. Carpenter JR, Kenward MG, White IR. Sensitivity analysis aftermultiple imputation under missing at random: a weightingapproach. Stat Methods Med Res. 2007;16:259-275.

27. Paul R, Campbell D, Gilbert K, Tsiouri I. Comparing spoken lan-guage treatments for minimally verbal preschoolers with autismspectrum disorders. J Autism Dev Disord. 2013;43:418-431.

28. Kaale A, Fagerland W, Martinsen EW, Smith L. Preschool-basedsocial-communication treatment for children with autism: 12-month follow-up of a randomized trial. J Am Acad Child AdolescPsychiatry. 2014;53:188-198.

29. Yoder P, Stone WL. Randomized comparison of two communi-cation interventions for preschoolers with autism spectrum dis-orders. J Consult Clin Psychol. 2006;74:426-435.

30. Kasari C, Gulsrud A, Freeman S, Paparella T, Hellemann G.Longitudinal follow up of children with autism receiving targetedinterventions on joint attention and play. J Am Acad Child Ado-lesc Psychiatry. 2012;51:487-495.

31. Schlosser RW, Wendt O. Effects of augmentative and alternativecommunication intervention on speech production in childrenwith autism: a systematic review. Am J Speech Lang Pathol. 2008;17:212-230.

32. Blischak DM, Lombardino LJ, Dyson AT. Use of speech-generating devices: in support of natural speech. AugmentAltern Commun. 2003;19:29-35.

OF THE AMERICAN ACADEMY OF CHILD & ADOLESCENT PSYCHIATRY


mailto:[email protected]

http://dx.doi.org/10.1016/j.jaac.2014.01.019


Documents

Communication Interventions for Minimally Verbal Children ... Kaiser et al 2014...sessments. Of the children, 60 met ADOS20 criteria for autism, and 1 met a criterion for ASD. Participants