Transcript of "National health statistics reports"
Number 65 n March 20, 2013Changes in Prevalence ofParent-reported Autism Spectrum Disorder inSchool-aged U.S. Children: 2007 to 2011–2012by Stephen J. Blumberg, Ph.D., Matthew D. Bramlett, Ph.D., National Center for Health Statistics;Michael D. Kogan, Ph.D., Maternal and Child Health Bureau;Laura A. Schieve, Ph.D., National Center on Birth Defects and Developmental Disabilities;Jessica R. Jones, M.P.H., and Michael C. Lu, M.D., M.P.H., Maternal and Child Health BureauAbstractObjectives—This report presents data on the prevalence of diagnosed autismspectrum disorder (ASD) as reported by parents of school-aged children (ages6–17 years) in 2011–2012. Prevalence changes from 2007 to 2011–2012 wereevaluated using cohort analyses that examine the consistency in the 2007 and2011–2012 estimates for children whose diagnoses could have been reported inboth surveys (i.e., those born in 1994–2005 and diagnosed in or before 2007).Data sources—Data were drawn from the 2007 and 2011–2012 NationalSurvey of Children’s Health (NSCH), which are independent nationallyrepresentative telephone surveys of households with children. The surveys wereconducted by the Centers for Disease Control and Prevention’s National Centerfor Health Statistics with funding and direction from the Health Resources andServices Administration’s Maternal and Child Health Bureau.Results—The prevalence of parent-reported ASD among children aged 6–17was 2.00% in 2011–2012, a significant increase from 2007 (1.16%). Themagnitude of the increase was greatest for boys and for adolescents aged 14–17.Cohort analyses revealed consistent estimates of both the prevalence of parent-reported ASD and autism severity ratings over time. Children who were firstdiagnosed in or after 2008 accounted for much of the observed prevalenceincrease among school-aged children (those aged 6–17). School-aged childrendiagnosed in or after 2008 were more likely to have milder ASD and less likelyto have severe ASD than those diagnosed in or before 2007.Conclusions—The results of the cohort analyses increase confidence thatdifferential survey measurement error over time was not a major contributor toobserved changes in the prevalence of parent-reported ASD. Rather, much of theprevalence increase from 2007 to 2011–2012 for school-aged children was theresult of diagnoses of children with previously unrecognized ASD.Keywords: autism prevalence • pervasive developmental disabilities • nationalestimates • State and Local Area Integrated Telephone SurveyIntroductionAutism spectrum disorder (ASD) isa set of complex neurodevelopmentdisorders that include autistic disorder,Asperger disorder, and pervasivedevelopmental disorder not otherwisespecified (1). Children who have ASDdisplay mild to severe impairments insocial interaction and communicationalong with restricted, repetitive, andstereotyped patterns of behaviors,interests, and activities. Diagnosis ofASD should be based on comprehensivebehavioral evaluations, makingdiagnostic assessment complex andtime-consuming.ASD symptoms typically can beidentified in children as young as 18months (2), and the American Academyof Pediatrics recommends developmentalscreening of all children by age 24months (3). Nevertheless, many childrenwith ASD—especially those with onlymild or limited speech delays—may notbe diagnosed until they are of schoolage, when parents become concernedabout an inability to make friends andteachers notice difficulties with peerinteractions (3). Formal diagnoses mayalso occur at this age because a nameddisability (such as ASD) is needed forU.S. DEPARTMENT OF HEALTH AND HUMAN SERVICESCenters for Disease Control and PreventionNational Center for Health Statistics
Page 2 National Health Statistics Reports n Number 65 n March 20, 2013school-aged children to qualify forspecial education services under theIndividuals with Disabilities EducationAct (4).The reported prevalence of ASD hasincreased in recent decades. Forexample, data from the Centers forDisease Control and Prevention’s (CDC)National Health Interview Survey(NHIS) revealed a nearly fourfoldincrease in parent-reported ASDbetween the 1997–1999 and 2006–2008surveillance periods (5), and CDC’sAutism and Developmental DisabilitiesMonitoring (ADDM) Network revealeda 78% increase in ASD prevalencebetween 2002 and 2008 (6).This report is the first to estimatethe change in prevalence of parent-reported ASD from 2007 to 2011–2012and is based on newly released datafrom the 2011–2012 National Survey ofChildren’s Health (NSCH). Estimatesare presented for the prevalence andseverity of parent-reported ASDdiagnoses for school-aged children(those aged 6–17 years). Detailedanalysis of ASD prevalence in thisgroup of children allowed forcomparisons with 2007 estimates for thesame birth cohort and an evaluation ofwhether observed changes in prevalencecould be due to changes in survey-basedmeasurement error.MethodsThe data were drawn from the 2007and 2011–2012 NSCH. NSCH is anationally representative surveyconducted by CDC’s National Center forHealth Statistics (NCHS) as a module ofthe State and Local Area IntegratedTelephone Survey, with direction andprincipal funding from the HealthResources and Services Administration’sMaternal and Child Health Bureau (7,8).In both 2007 and 2011–2012, NSCHwas fielded as a random-digit-dialtelephone survey of households withchildren aged 0–17 years in theUnited States; the 2011–2012 NSCHincluded both landlines and cell phones,whereas the 2007 NSCH includedlandlines only.Contacted households were screenedfor the presence of children, and onechild was randomly selected fromidentified households with children to bethe subject of the survey. Therespondent was a parent or guardian inthe household who was knowledgeableabout the child’s health. Theserespondents are referred to as ‘‘parents’’throughout this report. A total of 91,642NSCH 2007 interviews were completedfrom April 2007 through July 2008. Atotal of 95,677 NSCH 2011–2012interviews were completed fromFebruary 2011 through June 2012.The overall response rate for2011–2012 (23.0%) was lower than therate for 2007 (46.7%) primarily due tothe inclusion of cell-phone interviews in2011–2012. Nonresponse bias analysessuggest that, although the potential forbias cannot be ruled out, differencesbetween respondents andnonrespondents should not have a majorimpact on the conclusions in this report.Please see Technical Notes for details.This report is based on parent reportof ASD diagnoses, ASD severity, agewhen first diagnosed with ASD, andapproximate year of that first diagnosis.Survey question wording and thederivation of calculated variables aredescribed in the Technical Notes.Children classified as having ASD werethose with a parent report of ever beingtold by a doctor or other health careprovider that the child had ASD, and aparent report that the child currently hasASD. This measure of parent-reportedASD is analogous to measures of‘‘parent-reported ASD’’ (9) and ‘‘currentASD’’ (10) used in previously publishedreports on NSCH-based estimates ofASD prevalence and trends.Information about parent-reportedASD diagnosis was obtained for 63,967children aged 6–17 in 2007 and for65,556 in 2011–2012. The cohortanalyses include information aboutparent-reported ASD diagnosis for56,399 children aged 2–13 in 2007.A sampling weight was provided byNCHS with the data record for eachchild. This weight is based on theprobability of selection of the child’stelephone number, with adjustments forknown survey response biases andfurther adjustments to ensure thatweighted estimates match demographiccontrol totals from the U.S. CensusBureau’s American Community Survey.Estimates based on these weights,including all national estimates producedfor this report, are representative of thenoninstitutionalized population of U.S.children.For more information about NSCH,including its sample design, datacollection procedures, and questionnairecontent, visit http://www.cdc.gov/nchs/slaits/nsch.htm.ResultsBased on parent reports, theprevalence of diagnosed ASD in2011–2012 was estimated to be 2.00%for children aged 6–17. This prevalenceestimate (1 in 50) is significantly higherthan the estimate (1.16%, or 1 in 86) forchildren in that age group in 2007.As shown in Figure 1, statisticallysignificant increases in the prevalence ofparent-reported ASD were observed forall age groups (from 1.31% to 1.82%for ages 6–9; 1.45% to 2.39% for ages10–13; and 0.73% to 1.78% for ages14–17) and for boys (from 1.80% to3.23%). The increase for girls (from0.49% to 0.70%) was not statisticallysignificant. The increase in prevalencewas greater for boys (1.43 percentagepoints) than girls (0.21 percentagepoints) and greater for children aged14–17 (1.05 percentage points) than forchildren aged 6–9 (0.51 percentagepoints).In 2011–2012, school-aged boyswere more than 4 times as likely asschool-aged girls to have ASD (3.23%compared with 0.70%). Comparisons ofthe prevalence of ASD between thenarrower age groups (ages 6–9, 10–13,and 14–17) that compose the school-agegroup (ages 6–17) did not reveal anysignificant differences in 2011–2012. Incontrast, age-related differences wereobserved in 2007. At that time, childrenaged 14–17 were less likely to haveASD than children aged 10–13 (0.73%compared with 1.45%).Cohort analyses ofprevalenceThe observed prevalence increasesin parent-reported ASD could be due to
National Health Statistics Reports n Number 65 n March 20, 2013 Page 3SOURCE: CDC/NCHS, National Survey of Children’s Health, 2007 and 2011–2012.Percent0.00.51.01.52.02.53.03.5 2011–20122007Girls 6–17Boys 6–1714–1710–136–96–17Age in years1.161.311.450.731.78 1.800.492.001.822.393.230.70Figure 1. Percentage of children aged 6–17 years with parent-reported autism spectrum disorder, by age group and sex: United States,2007 and 2011–2012survey-based measurement changes overtime. One way to evaluate the potentialNOTES: ASD is autism spectrum disorder. NSCH is National Survey of Children’s Health.SOURCE: CDC/NCHS, National Survey of Children’s Health, 2007 and 2011–2012.Percent0.00.51.01.52.0NSCH 2011–2012,6–17 yearsNSCH 2007,2–13 yearsApproximate year ofASD diagnosis, in orafter 2008Approximate year ofASD diagnosis, in orbefore 20071.161.370.61Figure 2. Percentage of children born in 1994–2005 who have parent-reported autismspectrum disorder, by survey and approximate year when child was first diagnosed withASD: United States, 2007 and 2011–2012impact of measurement change is tocompare independent estimates from the2007 and 2011–2012 NSCH surveyyears for the same population (or birthcohort) of children for diagnoses thatcould have been reported in both surveyyears. If these prevalence estimates from2007 and 2011–2012 are identical, thiswould suggest that the observedprevalence increases are not due tosurvey-based measurement error.Children aged 6–17 in 2011–2012were born in 1994–2005(approximately) and would have beenabout ages 2–13 in 2007 when theprevious NSCH was conducted. Theestimated prevalence of parent-reportedASD for that birth cohort in 2007 was1.16% (Figure 2).The appropriate comparison groupfor those who were aged 2–13 in 2007would be those children aged 6–17 in2011–2012 whose ASD diagnosis couldhave been reported in 2007. That is, theappropriate comparison group in2011–2012 consists of children whowere first identified as having ASD inor before 2007. The estimated
Page 4 National Health Statistics Reports n Number 65 n March 20, 2013prevalence of parent-reported ASD forchildren aged 6–17 in 2011–2012 whoseASD was diagnosed in or before 2007was 1.37% (Figure 2). This estimate isstatistically indistinguishable (p > 0.10)from the estimate for that same birthcohort from the 2007 NSCH, suggestingthat survey-based measurement errorwas unlikely to have been a majorcontributor to the observed prevalenceincreases.The results shown in Figure 2 anddetailed in Table 1 also suggest that theobserved increase in ASD prevalencebetween 2007 and 2011–2012 amongchildren included in the 1994–2005 birthcohort is related to relatively recentdiagnoses. The 0.61% of children aged6–17 diagnosed with ASD in or after2008 largely accounts for the prevalenceincrease from 2007 to 2011–2012 forthis group.Table 1 provides estimates forsimilar comparisons within the threenarrower birth cohorts subsumed withinthe total 1994–2005 birth cohort group.For example, for children aged 6–9 in2011–2012 (the cohort born in 2002–2005), a comparison to the 2007estimate for children from the samebirth cohort (that is, children aged 2–5in 2007) illustrates that, for diagnosesreceived in or before 2007, the 2011–2012 ASD estimate (0.89%) was notsignificantly different from the 2007estimate (0.70%). Similar nonsignificantdifferences were observed for childrenaged 10–13 in 2011–2012 (1.68%compared with 1.31%) and for childrenaged 14–17 in 2011–2012 (1.53%compared with 1.45%).Table 1 also shows that, asexpected, relatively recent diagnoses(that is, those occurring in or after 2008)were more common for the cohorts ofchildren aged 6–9 (0.89%) and 10–13(0.71%) in 2011–2012 than for thecohort of children aged 14–17 in2011–2012 (0.24%). These relativelyrecent diagnoses largely account for theprevalence increases from 2007 to2011–2012 for these age groups.Children diagnosed in or after 2008represent 50.0%, 29.8%, and 13.5% ofchildren with parent-reported ASD whowere aged 6–9, 10–13, and 14–17 in2011–2012, respectively.Cohort analyses of severityCohort analyses can also be used toexamine the consistency of NSCHestimates of the distribution ofchildren’s ASD severity, as judged bytheir parents. The first two columns ofFigure 3 present a comparison of ASDseverity ratings between 2007 and2011–2012 for the cohort of childrenborn in 1994–2005, where the 2011–2012 estimates are restricted to thosechildren whose ASD diagnosis was firstreceived in or before 2007. Theconsistency was confirmed: Nodifferences in ASD severity wereobserved between children aged 2–13 in2007 and children aged 6–17 in2011–2012 who first received an ASDdiagnosis in or before 2007.The second and third columns ofFigure 3 show the distribution ofparent-reported ASD severity levels bythe approximate year when the ASDwas diagnosed. Relative to otherNOTES: ASD is autism spectrum disorder. NSCH is National Survey of Children’s Health.SOURCE: CDC/NCHS, National Survey of Children’s Health, 2007 and 2011–2012.Percent020406080100SevereModerateMildApproximate year of ASDdiagnosis, in or after 2008Approximate year of ASDdiagnosis, in or before 2007Approximate year of ASDdiagnosis, in or before 2007NSCH 2007, 2–13 years NSCH 2011–2012, 6–17 years48.7 49.558.335.1 33.634.816.3 16.96.9Figure 3. Percent distribution of parent-rated severity of autism spectrum disorder for children born in 1994–2005 who currently haveautism spectrum disorder, by survey and approximate year when child was first diagnosed with ASD: United States, 2007 and 2011–2012
National Health Statistics Reports n Number 65 n March 20, 2013 Page 5school-aged children with ASD in2011–2012, those children aged 6–17who were diagnosed in or after 2008were more likely to have mild ASD andless likely to have severe ASD, andstratum-adjusted Cochran-Mantel-Haenszel tests reveal that this trendtoward less severe ASD for more recentdiagnoses was statistically significant.Children aged 6–17 who were diagnosedin or after 2008 were less than one-halfas likely as children diagnosed in orbefore 2007 to have severe ASD (6.9%compared with 16.9%).Table 2 reveals that this trend ismost notable for older children. Amongchildren aged 14–17, almost none (1%)of the children with diagnoses in orafter 2008 were classified as havingsevere ASD, yet 18% of children whoseASD diagnosis was received in orbefore 2007 were classified as havingsevere ASD.Summary andDiscussionBetween 2007 and 2011–2012, theprevalence estimate for parent-reportedASD diagnoses among U.S. childrenaged 6–17 increased significantly, from1.16% to 2.00%. Increases wereobserved in all age groups, and amongboys aged 6–17.The increases in ASD prevalencereported here extend an ongoing trendobserved in the United States and otherdeveloped countries over the pastseveral decades (5,6,11). Assessment ofthe NSCH age-specific changes inparent-reported ASD prevalence revealstwo key components of this trend.First, the prevalence of parent-reported ASD for adolescents aged14–17 in 2011–2012 was greater thanthe prevalence for adolescents aged14–17 in 2007. The observed increase inthis cross-sectional comparison is duelargely to differences between the twobirth cohorts that are being compared,rather than a recent rise in newdiagnoses in adolescence. The birthcohort that was aged 14–17 in 2007(that is, children born in 1990–1993)had the lowest prevalence of parent-reported ASD among all birth cohorts in2007, but as young adults aged 18–21 in2011–2012, they were too old to be partof the 2011–2012 survey. The birthcohort that was aged 14–17 in 2011–2012 (children born in 1994–1997) hadthe highest prevalence of parent-reportedASD in 2007, when they were aged10–13. As adolescents aged 14–17 in2011–2012, this birth cohort continuedto have a similarly high prevalence ofparent-reported ASD, and thisprevalence was significantly higher thanthe observed prevalence for the birthcohort that aged out of the survey.Second, significant increases inparent-reported ASD prevalence wereobserved in both cross-sectional andcohort analyses for children aged 6–9and 10–13 in 2011–2012. While it is notpossible to definitively determine theunderlying reason for this trend, severalprevious studies that also examinedchanges over time in ASD prevalenceestimates within selected birth cohortssuggest that increasing recognition ofchildren with ASD has had an importantimpact (10,12,13). Evidence fromepidemiologic, genetic, and neuroscienceresearch points to the prenatal period asthe key exposure window for ASD riskfactors (14,15). If the causes of ASD arerelated to factors that exist prior to oroccur just after birth, it is unlikely thatchanges in ASD prevalence within thecohorts of children aged 6–13 in2011–2012 reflect ‘‘true’’ increases insusceptibility to the condition at theseages. Rather, changes in prevalence ofparent-reported ASD within these birthcohorts (children born in 1998–2005)likely reflect either changes in therecognition of ASD by healthprofessionals or survey-basedmeasurement changes over time.Closer examination of the parent-reported data confirm that ASDprevalence increases for the majority ofchildren were driven by new (post-2007)diagnoses rather than changes in surveyreliability, such as changes in howparents reported earlier (pre-2008)diagnoses. The change (or lack thereof)in how parents reported earlierdiagnoses was evaluated by comparing2007 and 2011–2012 estimates for asubgroup of the population limited tochildren represented in both surveys(that is, children from the same birthcohorts) and limited to those ASDdiagnoses that ostensibly could havebeen reported in both surveys (that is,diagnoses first received in or before2007). For this group of comparablechildren, the prevalence estimate andseverity distribution from the 2011–2012NSCH were not statistically differentfrom the prevalence estimate andseverity distribution from the 2007NSCH. The 2007 and 2011–2012 NSCHsamples were independent cross-sectional samples weighted to berepresentative of the U.S. population ofnoninstitutionalized children in therespective survey years, and the slightdifferences observed across surveys arewithin the range of possible samplingvariation.ASD prevalence increases due torelatively recent diagnoses (that is, thoseoccurring in or after 2008) were mostcommon for the cohorts of childrenaged 6–9 and 10–13 in 2011–2012, butwere also observed to a smaller degreefor children aged 14–17. Nearlyone-third of the school-aged childrenreported to have ASD in 2011–2012were reported to have been diagnosed inor after 2008. While this isunderstandable for many of the youngestchildren (those aged 6–9), 30% ofchildren aged 10–13 and 14% ofchildren aged 14–17 were firstdiagnosed in or after 2008 at age 7 orover, well beyond the age when ASDsigns and symptoms should be clearlynotable. More than one-half of thesechildren were classified by their parentsas having ‘‘mild’’ ASD, and very fewwere classified as having ‘‘severe’’ ASD.Together, these findings suggest that theincrease in prevalence of parent-reportedASD may have resulted from improvedascertainment of ASD by doctors andother health care professionals in recentyears, especially when the symptoms aremild. Changes in the ascertainment ofASD could occur because of changes inASD awareness among parents or healthcare professionals, increased access todiagnostic services, changes in howscreening tests or diagnostic criteria areused, or increased special educationplacements in the community.The analyses in this report haveseveral strengths, including a large,
Page 6 National Health Statistics Reports n Number 65 n March 20, 2013nationally representative sample, withsufficient sample size to explore ASDwithin selected strata defined by age orbirth cohort. Major limitations of theNSCH estimates are that they arepotentially subject to surveynonresponse bias and that parents’reports were not substantiated throughclinical evaluation or educationalrecords. Though the validity of parentreports may be uncertain, other studiesreporting previous NSCH autismestimates support their use. Forexample, parent-reported autismprevalence estimates from the 2003NSCH were highly concordant withNHIS estimates from the same timeperiod and were related as expected tosocial, emotional, and developmentaldifficulties experienced by children (16).Moreover, NSCH estimates from the2007 NSCH were comparable to theADDM Network estimates from thesame time period (6,9). The ADDMNetwork tracks children aged 8 residingin selected areas in the United States;ASD case finding is through expertclinical review of special education andmedical records from specialistproviders who serve children withdevelopmental disabilities. The 2011–2012 NSCH estimates cannot becompared in a similar manner becausethe autism-focused questions in NHISdiffer substantially from those in therecent NSCH, and ADDM data forcomparable years are not yet available.Data from the ADDM Networksupport the inference that there has beenan increase in ASD recognition bycommunity providers. While all childrendefined as ASD cases in ADDM havepreviously come to the attention of aspecial education or health care providerin their communities, not all of thesechildren had a previous diagnosis orschool classification specific to ASD.Each surveillance year, a percentage ofchildren who had clear documentationof autism traits in their records, but nota specific ASD diagnosis or educationalplacement, are nonetheless classified asASD cases by ADDM clinicalreviewers. This proportion of ‘‘ASD, butno previous diagnosis’’ cases hasdeclined steadily as total ASDprevalence rates have shown markedincreases (6).In conclusion, the consistency inindependent estimates of parent-reportedASD prevalence for children born in1994–2005 and diagnosed in or before2007 increases confidence that NSCHdata were not subject to differentialsurvey measurement error over time andtherefore can be used to monitorchanges in the prevalence of parent-reported ASD over time. Increases in theprevalence of parent-reported ASDcontinued through 2011–2012, and muchof the recent increase—especially forchildren aged 6–13—was the result ofdiagnoses of children with previouslyunrecognized ASD. A more detailedreport will further explore these findingswithin sociodemographic subgroups andwith further consideration of therelationship between health care systemfactors and ASD case ascertainment.References1. American Psychiatric Association.Diagnostic and Statistical Manual ofMental Disorders. 4th ed.Washington, DC: AmericanPsychiatric Association. 2000.2. Zwaigenbaum L, Bryson S, RogersT, Roberts W, Brian J, Szatmari P.Behavioral manifestations of autismin the first year of life. Int J DevNeurosci 23(2–3):143–52. 2005.3. Johnson CP, Myers SM.Identification and evaluation ofchildren with autism spectrumdisorders. Pediatrics 120(5):1183–215. 2007.4. Individuals with DisabilitiesEducation Act Amendments. Pub LNo. 101–476. 1990.5. Boyle CA, Boulet S, Schieve LA,Cohen RA, Blumberg SJ, Yeargin-Allsopp M, et al. Trends in theprevalence of developmentaldisabilities in US children, 1997–2008. Pediatrics 127(6):1034–42.2011.6. CDC. Prevalence of autism spectrumdisorders—Autism andDevelopmental DisabilitiesMonitoring Network, 14 sites,United States, 2008. MMWR 61(No.SS03):1–19. 2012.7. Blumberg SJ, Foster EB, Frasier AM,et al. Design and operation of theNational Survey of Children’sHealth, 2007. National Center forHealth Statistics. Vital Health Stat1(55). 2012.8. Centers for Disease Control andPrevention, National Center forHealth Statistics, State and LocalArea Integrated Telephone Survey.2011–2012 National Survey ofChildren’s Health frequently askedquestions. March 2013. Availablefrom: http://www.cdc.gov/nchs/slaits/nsch.htm.9. Kogan MD, Blumberg SJ, SchieveLA, Boyle CA, Perrin JM, GhandourRM, et al. Prevalence of parent-reported diagnosis of autismspectrum disorder among children inthe US, 2007. Pediatrics124(5):1395–403. 2009.10. Schieve LA, Rice C, Yeargin-AllsoppM, Boyle CA, Kogan MD, Drews C,Devine O. Parent-reported prevalenceof autism spectrum disorders inUS-born children: An assessment ofchanges within birth cohorts from the2003 to the 2007 National Survey ofChildren’s Health. Matern ChildHealth J 16(Suppl 1):S151–7. 2012.11. Fombonne E. Past and futureperspectives on autism epidemiology.In: SO Moldin, JLR Rubenstein(Eds.), Understanding Autism: FromBasic Neuroscience to Treatment25–47. Boca Raton, FL: Taylor andFrancis Group. 2006.12. Gurney JG, Fritz MS, Ness KK,Sievers P, Newschaffer CJ, ShapiroEG. Analysis of prevalence trends ofautism spectrum disorder inMinnesota. Arch Pediatr AdolescMed 157(7):622–7. 2003.13. Newschaffer CJ, Falb MD, GurneyJG. National autism prevalencetrends from United States specialeducation data. Pediatrics115(3):e277–82. 2005.14. Arndt TL, Stodgell CJ, Rodier PM.The teratology of autism. Int J DevNeurosci 23(2–3):189–99. 2005.15. Landrigan PJ. What causes autism?Exploring the environmentalcontribution. Curr Opin Pediatr22(2):219–25. 2010.16. CDC. Mental health in theUnited States: Parental report ofdiagnosed autism in children aged4–17 years—United States, 2003–2004. MMWR 55(17):481–6. 2006.17. Keeter S, Kennedy C, Dimock M,Best J, Craighill P. Gauging theimpact of growing nonresponse onestimates from a national RDD
National Health Statistics Reports n Number 65 n March 20, 2013 Page 7telephone survey. Public Opin Q70(5):759–79. 2006.18. Johnson TP, Wislar JS. Responserates and nonresponse errors insurveys. JAMA 307(17):1805–6.2012.19. Skalland BJ, Blumberg SJ.Nonresponse in the National Surveyof Children’s Health, 2007. NationalCenter for Health Statistics. VitalHealth Stat 2(156). 2012.20. Blumberg SJ, Luke JV. Wirelesssubstitution: Early release ofestimates from the National HealthInterview Survey, January–June2012. National Center for HealthStatistics. 2012. Available from:http://www.cdc.gov/nchs/nhis.htm.
Page 8 National Health Statistics Reports n Number 65 n March 20, 2013Table 1. Percentage of children with parent-reported autism spectrum disorder, by age cohort, approximate year of ASD diagnosis, andsurvey: United States, 2007 and 2011–2012NSCH 2007 NSCH 2011–2012Year of diagnosisApproximate birth year1Child agein yearsNumber ofchildrenwith ASDin sampleDiagnosisin orbefore2007Child agein yearsNumber ofchildrenwith ASDin sampleIn orbefore2007In orafter2008Percent ofchildrenwith ASDdiagnosedin orafter 20081994–2005 . . . . . . . . . . . . . . .2002–2005 . . . . . . . . . . . . . . .1998–2001 . . . . . . . . . . . . . . .1994–1997 . . . . . . . . . . . . . . .2–132–56–910–136721622592511.160.701.311.456–176–910–1314–171,3934405134401.370.891.681.530.610.890.710.2430.850.029.813.51Date of birth was not asked in the survey, so birth year was approximated from child’s age at interview.NOTES: ASD is autism spectrum disorder. Age cohort is defined by child’s age at interview. Year of ASD diagnosis is approximated from date of interview, child’s age at time of interview, andchild’s age when first diagnosed with ASD. See Technical Notes for more detail. The National Survey of Children’s Health (NSCH) is a cross-sectional survey. Because of the random samplingmethod, it is unlikely that any child was included in the survey in both 2007 and 2011–2012. Data are based on telephone interviews of parents or guardians of a sample of the noninstitutionalizedpopulation of U.S. children. All population estimates are based on sampling weights assigned to the data for each child.SOURCE: CDC/NCHS, National Survey of Children’s Health, 2007 and 2011–2012.
Page 10 National Health Statistics Reports n Number 65 n March 20, 2013Technical NotesResponse rates and analysisof nonresponseResponse rates provide one measureof the potential for nonresponse bias, thepossibility that the characteristics of thesample interviewed differ from thecharacteristics of the population ofinterest in some meaningful way. The2007 overall response rate was 46.7%and the 2011–2012 overall response ratewas 23.0%. The reduction in theresponse rate was due primarily to theinclusion of cell-phone interviews in2011–2012, which provided bettercoverage of the population of children atthe expense of lower response rates. Thelower overall response rates forcell-phone interviews were largely theresult of differences in the proportion oftelephone numbers that were notanswered and therefore provided noindication of whether the numberbelonged to a household. Thisnonresolution rate was 19.1% for thelandline sample and 51.4% for thecell-phone sample.The resolution rate (i.e., thecomplement of the nonresolution rate) isone component of the overall responserate, which is calculated as the productof the resolution rate, the screeningcompletion rate (i.e., the proportion ofknown households where the presenceor absence of children was successfullydetermined), and the interviewcompletion rate [i.e., the proportion ofknown households with children thatcompleted the National Survey ofChildren’s Health (NSCH) interview].Relative to the nonresolution rates, thelandline (54.1%) and cell-phone (41.2%)interview completion rates were moresimilar. For the landline sample, theinterview completion rate wascomparable to the 66.0% interviewcompletion rate from 2007.In general, differences betweenrespondents and nonrespondents areonly weakly associated with responserates for telephone surveys, andmeaningful nonresponse bias will existonly when the likelihood of surveyresponse is related to the surveyvariables of interest (17). Therefore,regardless of the response rate, it isrecommended that survey researchersevaluate the possible impact of suchdifferences on the quality of surveyestimates (18).Nonresponse bias analyses wereconducted with the 2011–2012 NSCHdata (8) using several recommendedapproaches (18). Analyses usingcontextual information in the samplingframe (such as sociodemographiccharacteristics of the geographic areasassociated with landline telephonenumbers) suggested that response biasesin these characteristics would have littleeffect on prevalence estimates ofparent-reported autism spectrum disorder(ASD). However, another approach—comparing ASD prevalence estimates forresponders who readily participated withestimates for those who required moreeffort to encourage participation—revealed the possibility of a small effect.Children whose parents participatedafter five or more calls (1.58%) wereless likely to have parent-reported ASDthan those whose parents participatedwith fewer calls (2.10%). If oneassumes that high-effort respondents aresimilar to nonrespondents, this findingwould suggest that nonresponse mayhave increased the 2011–2012 ASDprevalence estimate by up to0.16 percentage points. This assumptioncannot be validated with the currentdata, but if true, this potential bias isstill only one-fifth of the magnitude ofthe observed increase in parent-reportedASD over time (0.84 percentage points)and is smaller than potential samplingerror (the half-width of the 95%confidence interval was 0.20 percentagepoints).Similar nonresponse bias analysesare not available for prevalenceestimates from the 2007 NSCH, but it isnot unreasonable to expect that, as in2011–2012, parents of children withASD might have been somewhat morelikely to readily respond to the survey.Analyses for other health and healthcare variables from the 2007 surveyindicate that estimated biases in thesesurvey estimates were small (within therange of potential sampling error) andinconsistent in direction depending onthe approach used to estimate bias (19).Together, these results from the 2007and 2011–2012 surveys suggest that anydifferences between survey respondentsand survey nonrespondents should nothave had a major impact on theconclusions in this report; however, thepotential for such impact cannot becompletely ruled out.Impact of change in sampledesignThe 2011–2012 NSCH sampleincluded both landlines and cell phones,whereas the 2007 NSCH includedlandlines only. This change in sampledesign was necessitated by the increasedproportion of children living in cellphone-only households in 2011–2012(20); however, this change prompts thequestion of whether the observedincrease in prevalence of parent-reportedASD may have been related to theinclusion of cell-phone sample or to thenecessary changes to weightingprocedures to account for dual-framesampling. If ASD prevalence was higheramong children in cell-phone-onlyhouseholds than among children inhouseholds with landlines, then theiromission from the 2007 sample andinclusion in the 2011–2012 samplewould have inflated the observed ASDprevalence difference. However, theASD prevalence for children aged 6–17in cell-phone-only households (1.91%)in 2011–2012 was nonsignificantlylower than for children in landlinehouseholds (2.10%). Furthermore,unweighted analysis of the 2007 and2011–2012 data for children aged 6–17showed a 78% increase in ASDprevalence, similar to the observedincrease of 72% based on analyses withthe sampling weights. This suggests thatchanges in the weighting procedures dueto dual-frame sampling were unlikely tohave contributed to the observedincrease.Definition of termsAutism spectrum disorder (ASD)—Children with parent-reported ASD wereidentified based on parents’ orguardians’ responses to two questions.
National Health Statistics Reports n Number 65 n March 20, 2013 Page 11First, parents were asked if they hadever been told by a doctor or otherhealth care provider that their child had‘‘autism, Asperger’s disorder, pervasivedevelopmental disorder, or other autismspectrum disorder.’’ If parents respondedaffirmatively, they were asked: ‘‘Does[child] currently have autism or autismspectrum disorder?’’ Children classifiedas having ASD were those with a parentreport of (a) ever being told by a doctoror other health care provider that theirchild had ASD and (b) the childcurrently having ASD. These twoquestions were included in both the2007 and 2011–2012 NSCH.Age at diagnosis—In the 2011–2012NSCH, parents or guardians whoreported ever being told by a doctor orother health care provider that theirchild had ASD were asked: ‘‘How oldwas [child] when you were first told bya doctor or other health care providerthat [he/she] had autism or autismspectrum disorder?’’ This question waspreceded by a reminder about thedefinition of ASD: ‘‘Earlier you told methat [child] has been diagnosed withautism or an autism spectrum disorder,such as Asperger’s disorder or pervasivedevelopmental disorder.’’ The questionabout age at diagnosis was asked afterthe parent reported that the child hadever been diagnosed and before theparent was asked if the child currentlyhad ASD. This question did not appearon the 2007 NSCH.Approximate year of diagnosis—Forthe 2011–2012 NSCH, the year whenthe child was first identified as havingASD was approximated using data onthe child’s age at diagnosis, the child’sage at the time of the interview, and thedate of the interview. The differencebetween the age at interview (in years)and the age at diagnosis (in years) wascalculated. If the survey was completedin 2011 and the difference was 3 yearsor less, or if the survey was completedin 2012 and the difference was 4 yearsor less, then the year when the childwas first diagnosed was calculated as inor after 2008. Otherwise, the year thechild was first diagnosed was calculatedas in or before 2007.The 2007 NSCH did not include aquestion about age at diagnosis. Allchildren reported to have ASD in the2007 NSCH were assumed to have ayear of diagnosis that was in or before2007; that is, the year of diagnosis wasassumed to be in or before the year ofthe survey.Severity of ASD—Parents orguardians who reported that the childcurrently has ASD were asked toprovide a qualitative ranking of severity:‘‘Would you describe [his/her] autism orautism spectrum disorder as mild,moderate, or severe?’’
U.S. DEPARTMENT OFHEALTH & HUMAN SERVICESCenters for Disease Control and PreventionNational Center for Health Statistics3311 Toledo RoadHyattsville, MD 20782FIRST CLASS MAILPOSTAGE & FEES PAIDCDC/NCHSPERMIT NO. G-284OFFICIAL BUSINESSPENALTY FOR PRIVATE USE, $300National Health Statistics Reports n Number 65 n March 20, 2013AcknowledgmentNCHS thanks Benjamin J. Skalland and Kathleen B. Santos from NORC at the University of Chicago for providing the analysis of nonresponseincluded in this report.Suggested citationBlumberg SJ, Bramlett MD, Kogan MD, et al.Changes in prevalence of parent-reportedautism spectrum disorder in school-aged U.S.children: 2007 to 2011–2012. National healthstatistics reports; no 65. Hyattsville, MD:National Center for Health Statistics. 2013.Copyright informationAll material appearing in this report is in thepublic domain and may be reproduced orcopied without permission; citation as tosource, however, is appreciated.National Center for Health StatisticsEdward J. Sondik, Ph.D., DirectorJennifer H. Madans, Ph.D., Associate Directorfor ScienceDivision of Health Interview StatisticsJane F. Gentleman, Ph.D., DirectorFor free e-mail updates on NCHS publication releases, subscribe online at: http://www.cdc.gov/nchs/govdelivery.htm.For questions or general information about NCHS: Tel: 1–800–232–4636 • E-mail: email@example.com • Internet: http://www.cdc.gov/nchsDHHS Publication No. (PHS) 2013–1250 • CS238979