Automated Extraction of Diagnostic Criteria From Electronic Health Records for Autism Spectrum Disorders: Development, Evaluation, and Application.

BACKGROUND: Electronic health records (EHRs) bring many opportunities for information utilization. One such use is the surveillance conducted by the Centers for Disease Control and Prevention to track cases of autism spectrum disorder (ASD). This process currently comprises manual collection and review of EHRs of 4- and 8-year old children in 11 US states for the presence of ASD criteria. The work is time-consuming and expensive. OBJECTIVE: Our objective was to automatically extract from EHRs the description of behaviors noted by the clinicians in evidence of the diagnostic criteria in the Diagnostic and Statistical Manual of Mental Disorders (DSM). Previously, we reported on the classification of entire EHRs as ASD or not. In this work, we focus on the extraction of individual expressions of the different ASD criteria in the text. We intend to facilitate large-scale surveillance efforts for ASD and support analysis of changes over time as well as enable integration with other relevant data. METHODS: We developed a natural language processing (NLP) parser to extract expressions of 12 DSM criteria using 104 patterns and 92 lexicons (1787 terms). The parser is rule-based to enable precise extraction of the entities from the text. The entities themselves are encompassed in the EHRs as very diverse expressions of the diagnostic criteria written by different people at different times (clinicians, speech pathologists, among others). Due to the sparsity of the data, a rule-based approach is best suited until larger datasets can be generated for machine learning algorithms. RESULTS: We evaluated our rule-based parser and compared it with a machine learning baseline (decision tree). Using a test set of 6636 sentences (50 EHRs), we found that our parser achieved 76% precision, 43% recall (ie, sensitivity), and >99% specificity for criterion extraction. The performance was better for the rule-based approach than for the machine learning baseline (60% precision and 30% recall). For some individual criteria, precision was as high as 97% and recall 57%. Since precision was very high, we were assured that criteria were rarely assigned incorrectly, and our numbers presented a lower bound of their presence in EHRs. We then conducted a case study and parsed 4480 new EHRs covering 10 years of surveillance records from the Arizona Developmental Disabilities Surveillance Program. The social criteria (A1 criteria) showed the biggest change over the years. The communication criteria (A2 criteria) did not distinguish the ASD from the non-ASD records. Among behaviors and interests criteria (A3 criteria), 1 (A3b) was present with much greater frequency in the ASD than in the non-ASD EHRs. CONCLUSIONS: Our results demonstrate that NLP can support large-scale analysis useful for ASD surveillance and research. In the future, we intend to facilitate detailed analysis and integration of national datasets.

DSM Criteria that Best Differentiate Intellectual Disability from Autism Spectrum Disorder.

Clinical characteristics of autism spectrum disorder (ASD) and intellectual disability (ID) overlap, creating potential for diagnostic confusion. Diagnostic and statistical manual of mental disorders (DSM) criteria that best differentiate children with ID and some ASD features from those with comorbid ID and ASD were identified. Records-based surveillance of ASD among 8-year-old children across 14 US populations ascertained 2816 children with ID, with or without ASD. Area under the curve (AUC) was conducted to determine discriminatory power of DSM criteria. AUC analyses indicated that restricted interests or repetitive behaviors best differentiated between the two groups. A subset of 6 criteria focused on social interactions and stereotyped behaviors was most effective at differentiating the two groups (AUC of 0.923), while communication-related criteria were least discriminatory. Matching children with appropriate treatments requires differentiation between ID and ASD. Shifting to DSM-5 may improve differentiation with decreased emphasis on language-related behaviors.

Autism spectrum disorder prevalence and associations with air concentrations of lead, mercury, and arsenic.

Lead, mercury, and arsenic are neurotoxicants with known effects on neurodevelopment. Autism spectrum disorder (ASD) is a neurodevelopmental disorder apparent by early childhood. Using data on 4486 children with ASD residing in 2489 census tracts in five sites of the Centers for Disease Control and Prevention's Autism and Developmental Disabilities Monitoring (ADDM) Network, we used multi-level negative binomial models to investigate if ambient lead, mercury, and arsenic concentrations, as measured by the US Environmental Protection Agency National-Scale Air Toxics Assessment (EPA-NATA), were associated with ASD prevalence. In unadjusted analyses, ambient metal concentrations were negatively associated with ASD prevalence. After adjusting for confounding factors, tracts with air concentrations of lead in the highest quartile had significantly higher ASD prevalence than tracts with lead concentrations in the lowest quartile (prevalence ratio (PR) = 1.36; 95 '% CI: 1.18, 1.57). In addition, tracts with mercury concentrations above the 75th percentile (>1.7 ng/m(3)) and arsenic concentrations below the 75th percentile (≤0.13 ng/m(3)) had a significantly higher ASD prevalence (adjusted RR = 1.20; 95 % CI: 1.03, 1.40) compared to tracts with arsenic, lead, and mercury concentrations below the 75th percentile. Our results suggest a possible association between ambient lead concentrations and ASD prevalence and demonstrate that exposure to multiple metals may have synergistic effects on ASD prevalence.

Fetal alcohol syndrome among children aged 7-9 years - Arizona, Colorado, and New York, 2010.

Fetal alcohol syndrome (FAS) is a serious birth defect and developmental disorder caused by in utero exposure to alcohol. Assessment of the public health burden of FAS through surveillance has proven difficult; there is wide variation in reported prevalence depending on the study population and surveillance method. Generally, records-based birth prevalence studies report estimates of 0.2-1.5 per 1,000 live births, whereas studies that use in-person, expert assessment of school-aged children in a community report estimates of 6-9 per 1,000 population. The Fetal Alcohol Syndrome Surveillance Network II addressed some of the challenges in records-based ascertainment by assessing a period prevalence of FAS among children aged 7‒9 years in Arizona, Colorado, and New York. The prevalence across sites ranged from 0.3 to 0.8 per 1,000 children. Prevalence of FAS was highest among American Indian/Alaska Native children and lowest among Hispanic children. These estimates continue to be much lower than those obtained from studies using in-person, expert assessment. Factors that might contribute to this discrepancy include 1) inadequate recognition of the physical and behavioral characteristics of FAS by clinical care providers; 2) insufficient documentation of those characteristics in the medical record; and 3) failure to consider prenatal alcohol exposure with diagnoses of behavioral and learning problems. Addressing these factors through training of medical and allied health providers can lead to practice changes, ultimately increasing recognition and documentation of the characteristics of FAS.

Population-based prevalence of malaria among pregnant women in Enugu State, Nigeria: the Healthy Beginning Initiative.

BACKGROUND: Malaria adversely affects pregnant women and their fetuses or neonates. Estimates of the malaria burden in pregnant women based on health facilities often do not present a true picture of the problem due to the low proportion of women delivering at these facilities in malaria-endemic regions. METHODS: Data for this study were obtained from the Healthy Beginning Initiative using community-based sampling. Self-identified pregnant women between the ages of 17-45 years were recruited from churches in Enugu State, Nigeria. Malaria parasitaemia was classified as high and low based on the malaria plus system. RESULTS: Of the 2069 pregnant women for whom malaria parasitaemia levels were recorded, over 99 % tested positive for malaria parasitaemia, 62 % showed low parasitaemia and 38 % high parasitaemia. After controlling for confounding variables, odds for high parasitaemia were lower among those who had more people in the household (for every one person increase in a household, OR = 0.94, 95 % CI 0.89-0.99). CONCLUSION: Results of this study are consistent with hospital-based estimates of malaria during pregnancy in southeastern Nigeria. Based on the high prevalence of malaria parasitaemia in this sample, education on best practices to prevent malaria during pregnancy, and resources in support of these practices are urgently needed.

Methods for surveillance of fetal alcohol syndrome: The Fetal Alcohol Syndrome Surveillance Network II (FASSNetII) - Arizona, Colorado, New York, 2009 - 2014.

Surveillance of fetal alcohol syndrome (FAS) is important for monitoring the effects of prenatal alcohol exposure and describing the public health burden of this preventable disorder. Building on the infrastructure of the Fetal Alcohol Syndrome Surveillance Network (FASSNet, 1997-2002), in 2009 the Centers for Disease Control and Prevention awarded 5-year cooperative agreements to three states, Arizona, Colorado, and New York, to conduct population-based surveillance of FAS. The Fetal Alcohol Syndrome Surveillance Network II (FASSNetII, 2009-2014) developed a surveillance case definition based on three clinical criteria: characteristic facial features, central nervous system abnormalities, and growth deficiency. FASSNetII modified the FASSNet methods in three important ways: (1) estimation of a period prevalence rather than birth prevalence; (2) surveillance of FAS among school-age children (ages 7-9 years) to better document the central nervous system abnormalities that are not apparent at birth or during infancy; and (3) implementation of an expert clinical review of abstracted data for probable and confirmed cases classified through a computerized algorithm. FASSNetII abstracted data from multiple sources including birth records, medical records from child development centers or other specialty clinics, and administrative databases such as hospital discharge and Medicaid. One challenge of FASSNetII was its limited access to non-medical records. The FAS prevalence that could be estimated was that of the population identified through an encounter with the healthcare system. Clinical and public health programs that identify children affected by FAS provide critical information for targeting preventive, medical and educational services in this vulnerable population.

Sibling concordance for clinical features of Duchenne and Becker muscular dystrophies.

INTRODUCTION: The correlation of markers of disease severity among brothers with Duchenne or Becker muscular dystrophy has implications for clinical guidance and clinical trials. METHODS: Sibling pairs with Duchenne or Becker muscular dystrophy (n = 60) were compared for ages when they reached clinical milestones of disease progression, including ceased ambulation, scoliosis of ≥ 20°, and development of cardiomyopathy. RESULTS: The median age at which younger brothers reached each milestone, compared with their older brothers ranged from 25 months younger for development of cardiomyopathy to 2 months older for ceased ambulation. For each additional month of ambulation by the older brother, the hazard of ceased ambulation by the younger brother decreased by 4%. CONCLUSIONS: The ages when siblings reach clinical milestones of disease vary widely between siblings. However, the time to ceased ambulation for older brothers predicts the time to ceased ambulation for their younger brothers.

The downstream effects of COVID-19 on adolescent girls in the Peruvian Amazon: qualitative findings on how the pandemic affected education and reproductive health.

Due to COVID-19, schools were closed to mitigate disease spread. Past studies have shown that disruptions in education have unintended consequences for adolescents, including increasing their risk of school dropout, exploitation, gender-based violence, pregnancy and early unions. In Peru, the government closed schools from March 2020 to March 2022, declaring a national emergency that affected an estimated 8 million children. These closures may have unintended consequences, including increased adolescent pregnancy, particularly in Peru's rural, largely indigenous regions. Loreto, located in the Peruvian Amazon, has one of the highest adolescent pregnancy rates in the country and poor maternal and child health outcomes. The underlying causes may not be fully understood as data are limited, especially as we transition out of the pandemic. This qualitative study investigated the downstream effects of COVID-19 on adolescent education and reproductive health in Loreto's districts of Nauta and Parinari. In-depth interviews (n=41) were conducted with adolescents and community leaders. These were held in June 2022, 3 months after the reinstitution of in-person classes throughout Peru. Focus group discussions (FGDs) were also completed with community health workers and educators from the same study area in October 2022 to supplement our findings (3 FGDs, n=15). We observed that the economic, educational and health effects of the COVID-19 pandemic contributed to reduced contraceptive use, and increased school abandonment, early unions and adolescent pregnancy. The interplay between adolescent pregnancy and both early unions and school abandonment was bidirectional, with each acting as both a cause and consequence of the other.

Statewide county-level autism spectrum disorder prevalence estimates-seven U.S. states, 2018.

PURPOSE: Autism spectrum disorder (ASD) prevalence information is necessary for identifying community needs such as addressing disparities in identification and services. METHODS: Seven Autism and Developmental Disabilities Monitoring (ADDM) Network sites participated in a pilot project to link statewide health and education data to generate statewide and county-level prevalence estimates for a broader age range for their states for the first time. RESULTS: Statewide prevalence of ASD for ages 3-21 years in 2018 ranged from 1.5% in Tennessee and Wisconsin to 2.3% in Arizona. The median county-level prevalence of ASD was 1.4% of residents ages 3-21 years. More boys than girls had ASD at all sites, and prevalence was lower among non-Hispanic Black, Hispanic, Asian/Pacific Islander, and American Indian/Alaska Native residents compared to non-Hispanic White residents at most sites. ASD prevalence estimates for children aged 8 years were similar to 2018 ADDM Network estimates that used record review to provide more in-depth information, but showed greater variation for children aged 4 years. CONCLUSIONS: Linkage of statewide data sets provides less detailed but actionable local information when more resource-intensive methods are not possible.

Prevalence and Characteristics of Autism Spectrum Disorder Among Children Aged 8 Years - Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2020.

Problem/Condition: Autism spectrum disorder (ASD). Period Covered: 2020. Description of System: The Autism and Developmental Disabilities Monitoring (ADDM) Network is an active surveillance program that provides estimates of the prevalence of ASD among children aged 8 years. In 2020, there were 11 ADDM Network sites across the United States (Arizona, Arkansas, California, Georgia, Maryland, Minnesota, Missouri, New Jersey, Tennessee, Utah, and Wisconsin). To ascertain ASD among children aged 8 years, ADDM Network staff review and abstract developmental evaluations and records from community medical and educational service providers. A child met the case definition if their record documented 1) an ASD diagnostic statement in an evaluation, 2) a classification of ASD in special education, or 3) an ASD International Classification of Diseases (ICD) code. Results: For 2020, across all 11 ADDM sites, ASD prevalence per 1,000 children aged 8 years ranged from 23.1 in Maryland to 44.9 in California. The overall ASD prevalence was 27.6 per 1,000 (one in 36) children aged 8 years and was 3.8 times as prevalent among boys as among girls (43.0 versus 11.4). Overall, ASD prevalence was lower among non-Hispanic White children (24.3) and children of two or more races (22.9) than among non-Hispanic Black or African American (Black), Hispanic, and non-Hispanic Asian or Pacific Islander (A/PI) children (29.3, 31.6, and 33.4 respectively). ASD prevalence among non-Hispanic American Indian or Alaska Native (AI/AN) children (26.5) was similar to that of other racial and ethnic groups. ASD prevalence was associated with lower household income at three sites, with no association at the other sites.Across sites, the ASD prevalence per 1,000 children aged 8 years based exclusively on documented ASD diagnostic statements was 20.6 (range = 17.1 in Wisconsin to 35.4 in California). Of the 6,245 children who met the ASD case definition, 74.7% had a documented diagnostic statement of ASD, 65.2% had a documented ASD special education classification, 71.6% had a documented ASD ICD code, and 37.4% had all three types of ASD indicators. The median age of earliest known ASD diagnosis was 49 months and ranged from 36 months in California to 59 months in Minnesota.Among the 4,165 (66.7%) children with ASD with information on cognitive ability, 37.9% were classified as having an intellectual disability. Intellectual disability was present among 50.8% of Black, 41.5% of A/PI, 37.8% of two or more races, 34.9% of Hispanic, 34.8% of AI/AN, and 31.8% of White children with ASD. Overall, children with intellectual disability had earlier median ages of ASD diagnosis (43 months) than those without intellectual disability (53 months). Interpretation: For 2020, one in 36 children aged 8 years (approximately 4% of boys and 1% of girls) was estimated to have ASD. These estimates are higher than previous ADDM Network estimates during 2000-2018. For the first time among children aged 8 years, the prevalence of ASD was lower among White children than among other racial and ethnic groups, reversing the direction of racial and ethnic differences in ASD prevalence observed in the past. Black children with ASD were still more likely than White children with ASD to have a co-occurring intellectual disability. Public Health Action: The continued increase among children identified with ASD, particularly among non-White children and girls, highlights the need for enhanced infrastructure to provide equitable diagnostic, treatment, and support services for all children with ASD. Similar to previous reporting periods, findings varied considerably across network sites, indicating the need for additional research to understand the nature of such differences and potentially apply successful identification strategies across states.

Early Identification of Autism Spectrum Disorder Among Children Aged 4 Years - Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2020.

Problem/Condition: Autism spectrum disorder (ASD). Period Covered: 2020. Description of System: The Autism and Developmental Disabilities Monitoring Network is an active surveillance program that estimates prevalence and characteristics of ASD and monitors timing of ASD identification among children aged 4 and 8 years. In 2020, a total of 11 sites (located in Arizona, Arkansas, California, Georgia, Maryland, Minnesota, Missouri, New Jersey, Tennessee, Utah, and Wisconsin) conducted surveillance of ASD among children aged 4 and 8 years and suspected ASD among children aged 4 years. Surveillance included children who lived in the surveillance area at any time during 2020. Children were classified as having ASD if they ever received 1) an ASD diagnostic statement in an evaluation, 2) a special education classification of autism (eligibility), or 3) an ASD International Classification of Diseases (ICD) code (revisions 9 or 10). Children aged 4 years were classified as having suspected ASD if they did not meet the case definition for ASD but had a documented qualified professional's statement indicating a suspicion of ASD. This report focuses on children aged 4 years in 2020 compared with children aged 8 years in 2020. Results: For 2020, ASD prevalence among children aged 4 years varied across sites, from 12.7 per 1,000 children in Utah to 46.4 in California. The overall prevalence was 21.5 and was higher among boys than girls at every site. Compared with non-Hispanic White children, ASD prevalence was 1.8 times as high among Hispanic, 1.6 times as high among non-Hispanic Black, 1.4 times as high among Asian or Pacific Islander, and 1.2 times as high among multiracial children. Among the 58.3% of children aged 4 years with ASD and information on intellectual ability, 48.5% had an IQ score of ≤70 on their most recent IQ test or an examiner's statement of intellectual disability. Among children with a documented developmental evaluation, 78.0% were evaluated by age 36 months. Children aged 4 years had a higher cumulative incidence of ASD diagnosis or eligibility by age 48 months compared with children aged 8 years at all sites; risk ratios ranged from 1.3 in New Jersey and Utah to 2.0 in Tennessee. In the 6 months before the March 2020 COVID-19 pandemic declaration by the World Health Organization, there were 1,593 more evaluations and 1.89 more ASD identifications per 1,000 children aged 4 years than children aged 8 years received 4 years earlier. After the COVID-19 pandemic declaration, this pattern reversed: in the 6 months after pandemic onset, there were 217 fewer evaluations and 0.26 fewer identifications per 1,000 children aged 4 years than children aged 8 years received 4 years earlier. Patterns of evaluation and identification varied among sites, but there was not recovery to pre-COVID-19 pandemic levels by the end of 2020 at most sites or overall. For 2020, prevalence of suspected ASD ranged from 0.5 (California) to 10.4 (Arkansas) per 1,000 children aged 4 years, with an increase from 2018 at five sites (Arizona, Arkansas, Maryland, New Jersey, and Utah). Demographic and cognitive characteristics of children aged 4 years with suspected ASD were similar to children aged 4 years with ASD. Interpretation: A wide range of prevalence of ASD by age 4 years was observed, suggesting differences in early ASD identification practices among communities. At all sites, cumulative incidence of ASD by age 48 months among children aged 4 years was higher compared with children aged 8 years in 2020, indicating improvements in early identification of ASD. Higher numbers of evaluations and rates of identification were evident among children aged 4 years until the COVID-19 pandemic onset in 2020. Sustained lower levels of ASD evaluations and identification seen at a majority of sites after the pandemic onset could indicate disruptions in typical practices in evaluations and identification for health service providers and schools through the end of 2020. Sites with more recovery could indicate successful strategies to mitigate service interruption, such as pivoting to telehealth approaches for evaluation. Public Health Action: From 2016 through February of 2020, ASD evaluation and identification among the cohort of children aged 4 years was outpacing ASD evaluation and identification 4 years earlier (from 2012 until March 2016) among the cohort of children aged 8 years in 2020 . From 2016 to March 2020, ASD evaluation and identification among the cohort of children aged 4 years was outpacing that among children aged 8 years in 2020 from 2012 until March 2016. The disruptions in evaluation that coincided with the start of the COVID-19 pandemic and the increase in prevalence of suspected ASD in 2020 could have led to delays in ASD identification and interventions. Communities could evaluate the impact of these disruptions as children in affected cohorts age and consider strategies to mitigate service disruptions caused by future public health emergencies.

Interventions to increase early infant diagnosis of HIV infection: A systematic review and meta-analysis.

OBJECTIVES: Early infant diagnosis (EID) of HIV infection increases antiretroviral therapy initiation, which reduces pediatric HIV-related morbidity and mortality. This review aims to critically appraise the effects of interventions to increase uptake of early infant diagnosis. DESIGN: This is a systematic review and meta-analysis of interventions to increase the EID of HIV infection. We searched PubMed, EMBASE, CINAHL, and PsycINFO to identify eligible studies from inception of these databases to June 18, 2020. EID Uptake at 4-8 weeks of age was primary outcome assessed by the review. We conducted meta-analysis, using data from reports of included studies. The measure of the effect of dichotomous data was odds ratios (OR), with a 95% confidence interval. The grading of recommendations assessment, development, and evaluation (GRADE) approach was used to assess quality of evidence. SETTINGS: The review was not limited by time of publication or setting in which the studies conducted. PARTICIPANTS: HIV-exposed infants were participants. RESULTS: Database search and review of reference lists yielded 923 unique titles, out of which 16 studies involving 13,822 HIV exposed infants (HEI) were eligible for inclusion in the review. Included studies were published between 2014 and 2019 from Kenya, Nigeria, Uganda, South Africa, Zambia, and India. Of the 16 included studies, nine (experimental) and seven (observational) studies included had low to moderate risk of bias. The studies evaluated eHealth services (n = 6), service improvement (n = 4), service integration (n = 2), behavioral interventions (n = 3), and male partner involvement (n = 1). Overall, there was no evidence that any of the evaluated interventions, including eHealth, health systems improvements, integration of EID, conditional cash transfer, mother-to-mother support, or partner (male) involvement, was effective in increasing uptake of EID at 4-8 weeks of age. There was also no evidence that any intervention was effective in increasing HIV-infected infants' identification at 4-8 weeks of age. CONCLUSIONS: There is limited evidence to support the hypothesis that interventions implemented to increase uptake of EID were effective at 4-8 weeks of life. Further research is required to identify effective interventions that increase early infant diagnosis of HIV at 4-8 weeks of age. PROSPERO NUMBER: (CRD42020191738).

Progress and Disparities in Early Identification of Autism Spectrum Disorder: Autism and Developmental Disabilities Monitoring Network, 2002-2016.

OBJECTIVE: Early identification can improve outcomes for children with autism spectrum disorder (ASD). We sought to assess changes in early ASD identification over time and by co-occurring intellectual disability (ID) and race/ethnicity. METHOD: Using data for 2002-2016 from a biennial population-based ASD surveillance program among 8-year-old children in the United States, we defined identification as a child's earliest recorded ASD diagnosis or special education eligibility. Unidentified children had characteristics meeting the ASD surveillance case definition but no recorded identification by age 8 years. We calculated median age at identification among identified children, median age at identification including unidentified children, and cumulative incidence of identification by age 48 months. RESULTS: ASD identification by age 48 months was 4 times (95% CI: 3.6-4.3) as likely in 2016 as in 2002, with the largest increases among children without ID. Median age at ASD identification among identified children decreased 3 months during this time. Children of every race/ethnicity were more likely to be identified over time. There were racial disparities stratified by ID: in 2016, Black and Hispanic children without ID were less likely to be identified with ASD than were White children (both groups risk ratio: 0.7; 95% CI: 0.5-0.8), but Black children were 1.5 times (95% CI: 1.3-1.9) as likely as White children to be identified with ASD and ID. CONCLUSION: Substantial progress has been made to identify more children with ASD early, despite minimal decrease in median age at diagnosis. Considerable disparities remain in early ASD identification by race/ethnicity and co-occurring intellectual disability.

Early Identification of Autism Spectrum Disorder Among Children Aged 4 Years - Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2018.

PROBLEM/CONDITION: Autism spectrum disorder (ASD). PERIOD COVERED: 2018. DESCRIPTION OF SYSTEM: The Autism and Developmental Disabilities Monitoring Network is an active surveillance program that estimates ASD prevalence and monitors timing of ASD identification among children aged 4 and 8 years. This report focuses on children aged 4 years in 2018, who were born in 2014 and had a parent or guardian who lived in the surveillance area in one of 11 sites (Arizona, Arkansas, California, Georgia, Maryland, Minnesota, Missouri, New Jersey, Tennessee, Utah, and Wisconsin) at any time during 2018. Children were classified as having ASD if they ever received 1) an ASD diagnostic statement (diagnosis) in an evaluation, 2) a special education classification of ASD (eligibility), or 3) an ASD International Classification of Diseases (ICD) code. Suspected ASD also was tracked among children aged 4 years. Children who did not meet the case definition for ASD were classified as having suspected ASD if their records contained a qualified professional's statement indicating a suspicion of ASD. RESULTS: For 2018, the overall ASD prevalence was 17.0 per 1,000 (one in 59) children aged 4 years. Prevalence varied from 9.1 per 1,000 in Utah to 41.6 per 1,000 in California. At every site, prevalence was higher among boys than girls, with an overall male-to-female prevalence ratio of 3.4. Prevalence of ASD among children aged 4 years was lower among non-Hispanic White (White) children (12.9 per 1,000) than among non-Hispanic Black (Black) children (16.6 per 1,000), Hispanic children (21.1 per 1,000), and Asian/Pacific Islander (A/PI) children (22.7 per 1,000). Among children aged 4 years with ASD and information on intellectual ability, 52% met the surveillance case definition of co-occurring intellectual disability (intelligence quotient ≤70 or an examiner's statement of intellectual disability documented in an evaluation). Of children aged 4 years with ASD, 72% had a first evaluation at age ≤36 months. Stratified by census-tract-level median household income (MHI) tertile, a lower percentage of children with ASD and intellectual disability was evaluated by age 36 months in the low MHI tertile (72%) than in the high MHI tertile (84%). Cumulative incidence of ASD diagnosis or eligibility received by age 48 months was 1.5 times as high among children aged 4 years (13.6 per 1,000 children born in 2014) as among those aged 8 years (8.9 per 1,000 children born in 2010). Across MHI tertiles, higher cumulative incidence of ASD diagnosis or eligibility received by age 48 months was associated with lower MHI. Suspected ASD prevalence was 2.6 per 1,000 children aged 4 years, meaning for every six children with ASD, one child had suspected ASD. The combined prevalence of ASD and suspected ASD (19.7 per 1,000 children aged 4 years) was lower than ASD prevalence among children aged 8 years (23.0 per 1,000 children aged 8 years). INTERPRETATION: Groups with historically lower prevalence of ASD (non-White and lower MHI) had higher prevalence and cumulative incidence of ASD among children aged 4 years in 2018, suggesting progress in identification among these groups. However, a lower percentage of children with ASD and intellectual disability in the low MHI tertile were evaluated by age 36 months than in the high MHI group, indicating disparity in timely evaluation. Children aged 4 years had a higher cumulative incidence of diagnosis or eligibility by age 48 months compared with children aged 8 years, indicating improvement in early identification of ASD. The overall prevalence for children aged 4 years was less than children aged 8 years, even when prevalence of children suspected of having ASD by age 4 years is included. This finding suggests that many children identified after age 4 years do not have suspected ASD documented by age 48 months. PUBLIC HEALTH ACTION: Children born in 2014 were more likely to be identified with ASD by age 48 months than children born in 2010, indicating increased early identification. However, ASD identification among children aged 4 years varied by site, suggesting opportunities to examine developmental screening and diagnostic practices that promote earlier identification. Children aged 4 years also were more likely to have co-occurring intellectual disability than children aged 8 years, suggesting that improvement in the early identification and evaluation of developmental concerns outside of cognitive impairments is still needed. Improving early identification of ASD could lead to earlier receipt of evidence-based interventions and potentially improve developmental outcomes.

Prevalence and Characteristics of Autism Spectrum Disorder Among Children Aged 8 Years - Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2018.

PROBLEM/CONDITION: Autism spectrum disorder (ASD). PERIOD COVERED: 2018. DESCRIPTION OF SYSTEM: The Autism and Developmental Disabilities Monitoring (ADDM) Network conducts active surveillance of ASD. This report focuses on the prevalence and characteristics of ASD among children aged 8 years in 2018 whose parents or guardians lived in 11 ADDM Network sites in the United States (Arizona, Arkansas, California, Georgia, Maryland, Minnesota, Missouri, New Jersey, Tennessee, Utah, and Wisconsin). To ascertain ASD among children aged 8 years, ADDM Network staff review and abstract developmental evaluations and records from community medical and educational service providers. In 2018, children met the case definition if their records documented 1) an ASD diagnostic statement in an evaluation (diagnosis), 2) a special education classification of ASD (eligibility), or 3) an ASD International Classification of Diseases (ICD) code. RESULTS: For 2018, across all 11 ADDM sites, ASD prevalence per 1,000 children aged 8 years ranged from 16.5 in Missouri to 38.9 in California. The overall ASD prevalence was 23.0 per 1,000 (one in 44) children aged 8 years, and ASD was 4.2 times as prevalent among boys as among girls. Overall ASD prevalence was similar across racial and ethnic groups, except American Indian/Alaska Native children had higher ASD prevalence than non-Hispanic White (White) children (29.0 versus 21.2 per 1,000 children aged 8 years). At multiple sites, Hispanic children had lower ASD prevalence than White children (Arizona, Arkansas, Georgia, and Utah), and non-Hispanic Black (Black) children (Georgia and Minnesota). The associations between ASD prevalence and neighborhood-level median household income varied by site. Among the 5,058 children who met the ASD case definition, 75.8% had a diagnostic statement of ASD in an evaluation, 18.8% had an ASD special education classification or eligibility and no ASD diagnostic statement, and 5.4% had an ASD ICD code only. ASD prevalence per 1,000 children aged 8 years that was based exclusively on documented ASD diagnostic statements was 17.4 overall (range: 11.2 in Maryland to 29.9 in California). The median age of earliest known ASD diagnosis ranged from 36 months in California to 63 months in Minnesota. Among the 3,007 children with ASD and data on cognitive ability, 35.2% were classified as having an intelligence quotient (IQ) score ≤70. The percentages of children with ASD with IQ scores ≤70 were 49.8%, 33.1%, and 29.7% among Black, Hispanic, and White children, respectively. Overall, children with ASD and IQ scores ≤70 had earlier median ages of ASD diagnosis than children with ASD and IQ scores >70 (44 versus 53 months). INTERPRETATION: In 2018, one in 44 children aged 8 years was estimated to have ASD, and prevalence and median age of identification varied widely across sites. Whereas overall ASD prevalence was similar by race and ethnicity, at certain sites Hispanic children were less likely to be identified as having ASD than White or Black children. The higher proportion of Black children compared with White and Hispanic children classified as having intellectual disability was consistent with previous findings. PUBLIC HEALTH ACTION: The variability in ASD prevalence and community ASD identification practices among children with different racial, ethnic, and geographical characteristics highlights the importance of research into the causes of that variability and strategies to provide equitable access to developmental evaluations and services. These findings also underscore the need for enhanced infrastructure for diagnostic, treatment, and support services to meet the needs of all children.

Early Identification of Autism Spectrum Disorder Among Children Aged 4 Years - Early Autism and Developmental Disabilities Monitoring Network, Six Sites, United States, 2016.

PROBLEM/CONDITION: Autism spectrum disorder (ASD). PERIOD COVERED: 2016. DESCRIPTION OF SYSTEM: The Early Autism and Developmental Disabilities Monitoring (Early ADDM) Network, a subset of the overall ADDM Network, is an active surveillance program that estimates ASD prevalence and monitors early identification of ASD among children aged 4 years. Children included in surveillance year 2016 were born in 2012 and had a parent or guardian who lived in the surveillance area in Arizona, Colorado, Missouri, New Jersey, North Carolina, or Wisconsin, at any time during 2016. Children were identified from records of community sources including general pediatric health clinics, special education programs, and early intervention programs. Data from comprehensive evaluations performed by community professionals were abstracted and reviewed by trained clinicians using a standardized ASD surveillance case definition with criteria from the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5). RESULTS: In 2016, the overall ASD prevalence was 15.6 per 1,000 (one in 64) children aged 4 years for Early ADDM Network sites. Prevalence varied from 8.8 per 1,000 in Missouri to 25.3 per 1,000 in New Jersey. At every site, prevalence was higher among boys than among girls, with an overall male-to-female prevalence ratio of 3.5 (95% confidence interval [CI] = 3.1-4.1). Prevalence of ASD between non-Hispanic white (white) and non-Hispanic black (black) children was similar at each site (overall prevalence ratio: 0.9; 95% CI = 0.8-1.1). The prevalence of ASD using DSM-5 criteria was lower than the prevalence using Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (DSM-IV-TR) criteria at one of four sites that used criteria from both editions. Among sites where ≥60% of children aged 4 years had information about intellectual disability (intelligence quotient ≤70 or examiner's statement of intellectual disability documented in an evaluation), 53% of children with ASD had co-occurring intellectual disability. Of all children aged 4 years with ASD, 84% had a first evaluation at age ≤36 months and 71% of children who met the surveillance case definition had a previous ASD diagnosis from a community provider. Median age at first evaluation and diagnosis for this age group was 26 months and 33 months, respectively. Cumulative incidence of autism diagnoses received by age 48 months was higher for children aged 4 years than for those aged 8 years identified in Early ADDM Network surveillance areas in 2016. INTERPRETATION: In 2016, the overall prevalence of ASD in the Early ADDM Network using DSM-5 criteria (15.6 per 1,000 children aged 4 years) was higher than the 2014 estimate using DSM-5 criteria (14.1 per 1,000). Children born in 2012 had a higher cumulative incidence of ASD diagnoses by age 48 months compared with children born in 2008, which indicates more early identification of ASD in the younger group. The disparity in ASD prevalence has decreased between white and black children. Prevalence of co-occurring intellectual disability was higher than in 2014, suggesting children with intellectual disability continue to be identified at younger ages. More children received evaluations by age 36 months in 2016 than in 2014, which is consistent with Healthy People 2020 goals. Median age at earliest ASD diagnosis has not changed considerably since 2014. PUBLIC HEALTH ACTION: More children aged 4 years with ASD are being evaluated by age 36 months and diagnosed by age 48 months, but there is still room for improvement in early identification. Timely evaluation of children by community providers as soon as developmental concerns have been identified might result in earlier ASD diagnoses, earlier receipt of evidence-based interventions, and improved developmental outcomes.

Prevalence of Autism Spectrum Disorder Among Children Aged 8 Years - Autism and Developmental Disabilities Monitoring Network, 11 Sites, United States, 2016.

PROBLEM/CONDITION: Autism spectrum disorder (ASD). PERIOD COVERED: 2016. DESCRIPTION OF SYSTEM: The Autism and Developmental Disabilities Monitoring (ADDM) Network is an active surveillance program that provides estimates of the prevalence of ASD among children aged 8 years whose parents or guardians live in 11 ADDM Network sites in the United States (Arizona, Arkansas, Colorado, Georgia, Maryland, Minnesota, Missouri, New Jersey, North Carolina, Tennessee, and Wisconsin). Surveillance is conducted in two phases. The first phase involves review and abstraction of comprehensive evaluations that were completed by medical and educational service providers in the community. In the second phase, experienced clinicians who systematically review all abstracted information determine ASD case status. The case definition is based on ASD criteria described in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition. RESULTS: For 2016, across all 11 sites, ASD prevalence was 18.5 per 1,000 (one in 54) children aged 8 years, and ASD was 4.3 times as prevalent among boys as among girls. ASD prevalence varied by site, ranging from 13.1 (Colorado) to 31.4 (New Jersey). Prevalence estimates were approximately identical for non-Hispanic white (white), non-Hispanic black (black), and Asian/Pacific Islander children (18.5, 18.3, and 17.9, respectively) but lower for Hispanic children (15.4). Among children with ASD for whom data on intellectual or cognitive functioning were available, 33% were classified as having intellectual disability (intelligence quotient [IQ] ≤70); this percentage was higher among girls than boys (39% versus 32%) and among black and Hispanic than white children (47%, 36%, and 27%, respectively) [corrected]. Black children with ASD were less likely to have a first evaluation by age 36 months than were white children with ASD (40% versus 45%). The overall median age at earliest known ASD diagnosis (51 months) was similar by sex and racial and ethnic groups; however, black children with IQ ≤70 had a later median age at ASD diagnosis than white children with IQ ≤70 (48 months versus 42 months). INTERPRETATION: The prevalence of ASD varied considerably across sites and was higher than previous estimates since 2014. Although no overall difference in ASD prevalence between black and white children aged 8 years was observed, the disparities for black children persisted in early evaluation and diagnosis of ASD. Hispanic children also continue to be identified as having ASD less frequently than white or black children. PUBLIC HEALTH ACTION: These findings highlight the variability in the evaluation and detection of ASD across communities and between sociodemographic groups. Continued efforts are needed for early and equitable identification of ASD and timely enrollment in services.

Prevalence and Characteristics of Autism Spectrum Disorder Among Children Aged 4 Years - Early Autism and Developmental Disabilities Monitoring Network, Seven Sites, United States, 2010, 2012, and 2014.

PROBLEM/CONDITION: Autism spectrum disorder (ASD) is estimated to affect up to 3% of children in the United States. Public health surveillance for ASD among children aged 4 years provides information about trends in prevalence, characteristics of children with ASD, and progress made toward decreasing the age of identification of ASD so that evidence-based interventions can begin as early as possible. PERIOD COVERED: 2010, 2012, and 2014. DESCRIPTION OF SYSTEM: The Early Autism and Developmental Disabilities Monitoring (Early ADDM) Network is an active surveillance system that provides biennial estimates of the prevalence and characteristics of ASD among children aged 4 years whose parents or guardians lived within designated sites. During surveillance years 2010, 2012, or 2014, data were collected in seven sites: Arizona, Colorado, Missouri, New Jersey, North Carolina, Utah, and Wisconsin. The Early ADDM Network is a subset of the broader ADDM Network (which included 13 total sites over the same period) that has been conducting ASD surveillance among children aged 8 years since 2000. Each Early ADDM site covers a smaller geographic area than the broader ADDM Network. Early ADDM ASD surveillance is conducted in two phases using the same methods and project staff members as the ADDM Network. The first phase consists of reviewing and abstracting data from children's records, including comprehensive evaluations performed by community professionals. Sources for these evaluations include general pediatric health clinics and specialized programs for children with developmental disabilities. In addition, special education records (for children aged ≥3 years) were reviewed for Arizona, Colorado, New Jersey, North Carolina, and Utah, and early intervention records (for children aged 0 to <3 years) were reviewed for New Jersey, North Carolina, Utah, and Wisconsin; in Wisconsin, early intervention records were reviewed for 2014 only. The second phase involves a review of the abstracted evaluations by trained clinicians using a standardized case definition and method. A child is considered to meet the surveillance case definition for ASD if one or more comprehensive evaluations of that child completed by a qualified professional describes behaviors consistent with the Diagnostic and Statistical Manual of Mental Disorders, 4th Edition, Text Revision (DSM-IV-TR) diagnostic criteria for any of the following conditions: autistic disorder, pervasive developmental disorder-not otherwise specified (PDD-NOS, including atypical autism), or Asperger disorder (2010, 2012, and 2014). For 2014 only, prevalence estimates based on surveillance case definitions according to DSM-IV-TR and the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5) were compared. This report provides estimates of overall ASD prevalence and prevalence by sex and race/ethnicity; characteristics of children aged 4 years with ASD, including age at first developmental evaluation, age at ASD diagnosis, and cognitive function; and trends in ASD prevalence and characteristics among Early ADDM sites with data for all 3 surveillance years (2010, 2012, and 2014), including comparisons with children aged 8 years living in the same geographic area. Analyses of time trends in ASD prevalence are restricted to the three sites that contributed data for all 3 surveillance years with consistent data sources (Arizona, Missouri, and New Jersey). RESULTS: The overall ASD prevalence was 13.4 per 1,000 children aged 4 years in 2010, 15.3 in 2012, and 17.0 in 2014 for Early ADDM sites with data for the specific years. ASD prevalence was determined using a surveillance case definition based on DSM-IV-TR. Within each surveillance year, ASD prevalence among children aged 4 years varied across surveillance sites and was lowest each year for Missouri (8.5, 8.1, and 9.6 per 1,000, for 2010, 2012, and 2014, respectively) and highest each year for New Jersey (19.7, 22.1, and 28.4 per 1,000, for the same years, respectively). Aggregated prevalence estimates were higher for sites that reviewed education and health care records than for sites that reviewed only health care records. Among all participating sites and years, ASD prevalence among children aged 4 years was consistently higher among boys than girls; prevalence ratios ranged from 2.6 (Arizona and Wisconsin in 2010) to 5.2 boys per one girl (Colorado in 2014). In 2010, ASD prevalence was higher among non-Hispanic white children than among Hispanic children in Arizona and non-Hispanic black children in Missouri; no other differences were observed by race/ethnicity. Among four sites with ≥60% data on cognitive test scores (Arizona, New Jersey, North Carolina, and Utah), the frequency of co-occurring intellectual disabilities was significantly higher among children aged 4 years than among those aged 8 years for each site in each surveillance year except Arizona in 2010. The percentage of children with ASD who had a first evaluation by age 36 months ranged from 48.8% in Missouri in 2012 to 88.9% in Wisconsin in 2014. The percentage of children with a previous ASD diagnosis from a community provider varied by site, ranging from 43.0% for Arizona in 2012 to 86.5% for Missouri in 2012. The median age at earliest known ASD diagnosis varied from 28 months in North Carolina in 2014 to 39.0 months in Missouri and Wisconsin in 2012. In 2014, the ASD prevalence based on the DSM-IV-TR case definition was 20% higher than the prevalence based on the DSM-5 (17.0 versus 14.1 per 1,000, respectively). Trends in ASD prevalence and characteristics among children aged 4 years during the study period were assessed for the three sites with data for all 3 years and consistent data sources (Arizona, Missouri, and New Jersey) using the DSM-IV-TR case definition; prevalence was higher in 2014 than in 2010 among children aged 4 years in New Jersey and was stable in Arizona and Missouri. In Missouri, ASD prevalence was higher among children aged 8 years than among children aged 4 years. The percentage of children with ASD who had a comprehensive evaluation by age 36 months was stable in Arizona and Missouri and decreased in New Jersey. In the three sites, no change occurred in the age at earliest known ASD diagnosis during 2010-2014. INTERPRETATION: The findings suggest that ASD prevalence among children aged 4 years was higher in 2014 than in 2010 in one site and remained stable in others. Among children with ASD, the frequency of cognitive impairment was higher among children aged 4 years than among those aged 8 years and suggests that surveillance at age 4 years might more often include children with more severe symptoms or those with co-occurring conditions such as intellectual disability. In the sites with data for all years and consistent data sources, no change in the age at earliest known ASD diagnosis was found, and children received their first developmental evaluation at the same or a later age in 2014 compared with 2010. Delays in the initiation of a first developmental evaluation might adversely affect children by delaying access to treatment and special services that can improve outcomes for children with ASD. PUBLIC HEALTH ACTION: Efforts to increase awareness of ASD and improve the identification of ASD by community providers can facilitate early diagnosis of children with ASD. Heterogeneity of results across sites suggests that community-level differences in evaluation and diagnostic services as well as access to data sources might affect estimates of ASD prevalence and age of identification. Continuing improvements in providing developmental evaluations to children as soon as developmental concerns are identified might result in earlier ASD diagnoses and earlier receipt of services, which might improve developmental outcomes.

Optimizing Corpus Creation for Training Word Embedding in Low Resource Domains: A Case Study in Autism Spectrum Disorder (ASD).

Automating the extraction of behavioral criteria indicative of Autism Spectrum Disorder (ASD) in electronic health records (EHRs) can contribute significantly to the effort to monitor the condition. Word embedding algorithms such as Word2Vec can encode semantic meanings of words in vectors and assist in automated vocabulary discovery from EHRs. However, text available for training word embeddings for ASD is miniscule compared to the billions of tokens typically used. We evaluate the importance of corpus specificity versus size and hypothesize that for specific domains small corpora can generate excellent word embeddings. We custom-built 6 ASD-themed corpora (N=4482), using ASD EHRs and abstracts from PubMed (N=39K) and PsychInfo (N=69K) and evaluated them. We were able to generate the most useful 200-dimension embeddings based on the small ASD EHR data. Due to diversity in its vocabulary, the abstract-based embeddings generated fewer related terms and saw minimal improvement when the size of the corpus increased.

Characterizing Health Disparities in the Age of Autism Diagnosis in a Study of 8-Year-Old Children.

The diagnosis of autism spectrum disorder (ASD) is often delayed from the time of noted concerns to the actual diagnosis. The current study used child- and family-level factors to identify homogeneous classes in a surveillance-based sample (n = 2303) of 8-year-old children with ASD. Using latent class analysis, a 5-class model emerged and the class memberships were examined in relation to the child's median age at ASD diagnosis. Class 3, with known language delays and a high advantage socioeconomically had the lowest age of ASD diagnosis (46.74 months) in comparison to Classes 1 (64.99 months), 4 (58.14 months), and 5 (69.78 months) in this sample. Findings demonstrate sociodemographic and developmental disparities related to the age at ASD diagnosis.