Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Finding Effective Screening Instruments for Autism Using Bayes Theorem

Finding Effective Screening Instruments for Autism Using Bayes Theorem Several screening instruments for the early detection of autism in young children have been developed.1-4 The usefulness of these instruments in clinical practice remains to be established, because many children need to be screened to correctly identify one child with autism. According to Bayes theorem, the chance that a disease is truly present (posttest odds) depends on both the prevalence of the disease (pretest odds) and the properties of the test (likelihood ratio). Thus, the outcome of screening tests depends on the chance of the disease being present in the population to which the test is applied. In case of rare diseases, screening instruments need to be very accurate to be useful. We assessed whether the screening tests for autism are accurate enough to overcome the low prevalence of autism spectrum disorders. Methods We conducted a systematic search of the literature on screening instruments for autism and included articles from peer-reviewed journals. Likelihood ratios were calculated from sensitivity and specificity. Estimates of sensitivity and specificity of the Early Screening for Autistic Traits (ESAT) were calculated with the prevalence numbers of autism spectrum disorders reported by Baird et al.1 Results The positive likelihood ratio found for the Checklist for Autism in Toddlers (CHAT) is 16,1 while the estimated likelihood ratio for the ESAT is 67.4 Estimates of the test properties of the Modified CHAT and the Social Communication Questionnaire in an unselected population are not yet available (Table).2,3 The ESAT's higher likelihood ratio comes with a loss of sensitivity, compared with the CHAT. When a higher threshold for the CHAT is used, the test properties of the CHAT and ESAT resemble each other. Table. View LargeDownload Available Screening Instruments for Early Detection of Autism Although a likelihood ratio greater than 10 is considered to substantially improve the likelihood that a disease is present (for comparison, the likelihood ratio of auscultation for detecting pneumonia is 3.7),5 the likelihood of a true positive result is equally influenced by the pretest odds. Because the prevalence of autism in the general unselected population is very low (6 cases per 1000 individuals for autism spectrum disorders, including autism, pervasive developmental disorders not otherwise specified, and Asperger syndrome),6 the CHAT's likelihood ratio of 16 gives a less than 10% probability that a child with a positive outcome indeed has autism. Thus, only 1 in every 10 children referred for further psychiatric evaluation will be diagnosed with autism. The estimated likelihood ratio of the ESAT gives a 30% chance that a screen-positive child has autism. Despite the high likelihood ratio for autism in both tests, screening for autism in the unselected population is inefficient because of the low prevalence of autism. All of the children who were screen positive on the ESAT, however, did have a developmental disorder. Thus, the ESAT's likelihood ratio for general developmental problems, instead of autism, is very high. Comment Presumably, the test characteristics of screening instruments for autism will not improve sufficiently to overcome the small pretest odds. One possibility to obtain more acceptable posttest odds is to increase the pretest odds. This can be done by applying screening instruments only to selected children that are found to have a deviant developmental path in routine developmental surveillance or have high-risk status by other means, such as being a younger sibling of a child diagnosed with autism.4 Such a 2-stage model could include nonspecific developmental deviations (such as abnormality or delay in linguistic, social, or motor development; presence of sensory handicaps; or the parents' feeling that something is wrong) that are followed by a specific screening instrument for autism. Correspondence: Dr Groen, Child and Adolescent Psychiatry, University Medical Centre, PO Box 9101, Nijmegen 6500 HB, the Netherlands (w.groen@psy.umcn.nl). Author Contributions:Study concept and design: Groen, Swinkels, van der Gaag, and Buitelaar. Analysis and interpretation of data: Groen and Swinkels. Drafting of the manuscript: Groen, Swinkels, van der Gaag, and Buitelaar. Critical revision of the manuscript for important intellectual content: Swinkels and Buitelaar. Statistical analysis: Groen and Swinkels. Study supervision: Swinkels, van der Gaag, and Buitelaar. Financial Disclosure: None reported. References 1. Baird GCharman TBaron-Cohen S et al. A screening instrument for autism at 18 months of age: a 6-year follow-up study. J Am Acad Child Adolesc Psychiatry 2000;39694- 702PubMedGoogle ScholarCrossref 2. Robins DLFein DBarton MGreen J The Modified Checklist for Autism in Toddlers: an initial study investigating the early detection of autism and pervasive developmental disorders. J Autism Dev Disord 2001;31131- 144PubMedGoogle ScholarCrossref 3. Berument SKRutter MLord CPickles ABailey A Autism screening questionnaire: diagnostic validity. Br J Psychiatry 1999;175444- 451PubMedGoogle ScholarCrossref 4. Dietz CSwinkels SVan Daalen EVan Engeland HBuitelaar J Screening for autistic spectrum disorder in children aged 14-15 months: II, population screening with the Early Screening of Autistic Traits Questionnaire (ESAT), design and general findings. J Autism Dev Disord 2006;36713- 722PubMedGoogle ScholarCrossref 5. Melbye HStraume BAasebo UDale K Diagnosis of pneumonia in adults in general practice: relative importance of typical symptoms and abnormal chest signs evaluated against a radiographic reference standard. Scand J Prim Health Care 1992;10226- 233PubMedGoogle ScholarCrossref 6. Fombonne E The prevalence of autism. JAMA 2003;28987- 89PubMedGoogle ScholarCrossref http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Archives of Pediatrics & Adolescent Medicine American Medical Association

Finding Effective Screening Instruments for Autism Using Bayes Theorem

Download PDF
2 pages

Loading next page...
 
/lp/american-medical-association/finding-effective-screening-instruments-for-autism-using-bayes-theorem-xKX4LywX29

References (6)

Publisher
American Medical Association
Copyright
Copyright © 2007 American Medical Association. All Rights Reserved.
ISSN
1072-4710
DOI
10.1001/archpedi.161.4.415
pmid
17404143
Publisher site
See Article on Publisher Site

Abstract

Several screening instruments for the early detection of autism in young children have been developed.1-4 The usefulness of these instruments in clinical practice remains to be established, because many children need to be screened to correctly identify one child with autism. According to Bayes theorem, the chance that a disease is truly present (posttest odds) depends on both the prevalence of the disease (pretest odds) and the properties of the test (likelihood ratio). Thus, the outcome of screening tests depends on the chance of the disease being present in the population to which the test is applied. In case of rare diseases, screening instruments need to be very accurate to be useful. We assessed whether the screening tests for autism are accurate enough to overcome the low prevalence of autism spectrum disorders. Methods We conducted a systematic search of the literature on screening instruments for autism and included articles from peer-reviewed journals. Likelihood ratios were calculated from sensitivity and specificity. Estimates of sensitivity and specificity of the Early Screening for Autistic Traits (ESAT) were calculated with the prevalence numbers of autism spectrum disorders reported by Baird et al.1 Results The positive likelihood ratio found for the Checklist for Autism in Toddlers (CHAT) is 16,1 while the estimated likelihood ratio for the ESAT is 67.4 Estimates of the test properties of the Modified CHAT and the Social Communication Questionnaire in an unselected population are not yet available (Table).2,3 The ESAT's higher likelihood ratio comes with a loss of sensitivity, compared with the CHAT. When a higher threshold for the CHAT is used, the test properties of the CHAT and ESAT resemble each other. Table. View LargeDownload Available Screening Instruments for Early Detection of Autism Although a likelihood ratio greater than 10 is considered to substantially improve the likelihood that a disease is present (for comparison, the likelihood ratio of auscultation for detecting pneumonia is 3.7),5 the likelihood of a true positive result is equally influenced by the pretest odds. Because the prevalence of autism in the general unselected population is very low (6 cases per 1000 individuals for autism spectrum disorders, including autism, pervasive developmental disorders not otherwise specified, and Asperger syndrome),6 the CHAT's likelihood ratio of 16 gives a less than 10% probability that a child with a positive outcome indeed has autism. Thus, only 1 in every 10 children referred for further psychiatric evaluation will be diagnosed with autism. The estimated likelihood ratio of the ESAT gives a 30% chance that a screen-positive child has autism. Despite the high likelihood ratio for autism in both tests, screening for autism in the unselected population is inefficient because of the low prevalence of autism. All of the children who were screen positive on the ESAT, however, did have a developmental disorder. Thus, the ESAT's likelihood ratio for general developmental problems, instead of autism, is very high. Comment Presumably, the test characteristics of screening instruments for autism will not improve sufficiently to overcome the small pretest odds. One possibility to obtain more acceptable posttest odds is to increase the pretest odds. This can be done by applying screening instruments only to selected children that are found to have a deviant developmental path in routine developmental surveillance or have high-risk status by other means, such as being a younger sibling of a child diagnosed with autism.4 Such a 2-stage model could include nonspecific developmental deviations (such as abnormality or delay in linguistic, social, or motor development; presence of sensory handicaps; or the parents' feeling that something is wrong) that are followed by a specific screening instrument for autism. Correspondence: Dr Groen, Child and Adolescent Psychiatry, University Medical Centre, PO Box 9101, Nijmegen 6500 HB, the Netherlands (w.groen@psy.umcn.nl). Author Contributions:Study concept and design: Groen, Swinkels, van der Gaag, and Buitelaar. Analysis and interpretation of data: Groen and Swinkels. Drafting of the manuscript: Groen, Swinkels, van der Gaag, and Buitelaar. Critical revision of the manuscript for important intellectual content: Swinkels and Buitelaar. Statistical analysis: Groen and Swinkels. Study supervision: Swinkels, van der Gaag, and Buitelaar. Financial Disclosure: None reported. References 1. Baird GCharman TBaron-Cohen S et al. A screening instrument for autism at 18 months of age: a 6-year follow-up study. J Am Acad Child Adolesc Psychiatry 2000;39694- 702PubMedGoogle ScholarCrossref 2. Robins DLFein DBarton MGreen J The Modified Checklist for Autism in Toddlers: an initial study investigating the early detection of autism and pervasive developmental disorders. J Autism Dev Disord 2001;31131- 144PubMedGoogle ScholarCrossref 3. Berument SKRutter MLord CPickles ABailey A Autism screening questionnaire: diagnostic validity. Br J Psychiatry 1999;175444- 451PubMedGoogle ScholarCrossref 4. Dietz CSwinkels SVan Daalen EVan Engeland HBuitelaar J Screening for autistic spectrum disorder in children aged 14-15 months: II, population screening with the Early Screening of Autistic Traits Questionnaire (ESAT), design and general findings. J Autism Dev Disord 2006;36713- 722PubMedGoogle ScholarCrossref 5. Melbye HStraume BAasebo UDale K Diagnosis of pneumonia in adults in general practice: relative importance of typical symptoms and abnormal chest signs evaluated against a radiographic reference standard. Scand J Prim Health Care 1992;10226- 233PubMedGoogle ScholarCrossref 6. Fombonne E The prevalence of autism. JAMA 2003;28987- 89PubMedGoogle ScholarCrossref

Journal

Archives of Pediatrics & Adolescent Medicine – American Medical Association

Published: Apr 1, 2007

There are no references for this article.