Accuracy of the Denver-I! in Developmental Screening

Transcription

1 Accuracy of the Denver-I! in Developmental Screening Frances Page Glascoe, PhD*; Karen E. Byrne, MA*; Linda G. Ashford, PhD*; Katherine L. Johnson, MAX; Bernard Chang; and Bryan Strickland ABSTRACT. One of the oldest and best known developmental screening tests was recently restandardized and revised as the Denver-Il. Because it was published without evidence of its accuracy, the present study was undertaken with 104 children between 3 and 72 months of age attending one of five day-care centers. To determine the presence of developmental problems, children were administered individual measures of intelligence, speech-language, achievement, and adaptive behavior. A second psychological examiner, blind to the outcome of the diagnostic battery, administered the Denver-H. Developmental problems including language impairments, learning disabilities, mild mental retardation, and/or functional developmental delay were found in 17% of the children. The Denver-I! identified correctly 83% and thus had high rates of sensitivity. However, more than half the children with normal development also received abnormal, questionable, or untestable Denver-Il scores. Thus the test had limited specificity (43%) and a high overreferral rate. The alternative scoring method, categorizing questionable/untestable scores as normal, caused sensitivity to drop to 56% although specificity rose to 80%. Since neither scoring method produced acceptable levels of accuracy, an effort was made to locate the sources of accuracy and inaccuracy within the test. Only items in the language domain were modestly helpfiil in discriminating children with and without difficulties. The findings suggest that the authors of the Denver- II need to engage in further development of the instrument including revising scoring criteria and item placement in relation to children s ages. In the interim, test users should employ screening tests which are more accurate such as the Minnesota Inventories or the Battelle Developmental Inventory Screening Test. Pediatrics 1992;89: ; developmental screening, child development, early identification. The Denver Developmental Screening Test is one of the oldest and best known brief measures of devebopment. First published in 1967, it is widely used nationally and internationally and has been standardized in 15 different countries. The test has also been the focus of much research, and a number of studies have been critical of its sensitivity to a variety of mild handicapping conditions. 5 As a consequence, the test has been substantially revised, restandardized, and renamed, the Denver-Il.6 From the Child Development Center, Department of Pediatrics, Vanderbilt University School of Medicine; Department of Psychology, George Peabody College for Teachers of Vanderbilt University; and Vanderbilt University School of Medicine, Nashville, IN. Received for publication May 17, 1991; accepted Jan 28, Reprint requests to (F.P.G.) Child Development Center, Dept of Pediatrics, 2100 Pierce Ave. Nashville, IN PEDIATRICS (ISSN ). Copyright 1992 by the American Academy of Pediatrics. The Denver-Il differs from its predecessor by the addition of 20 new items, most of which tap expressive language and articulation skills. Five separate items enable ratings of behavior and speech intelligibility. Sectors, now known as domains, continue to be cabled Personal-Social, Fine Motor-Adaptive, Language, and Gross Motor. However, there is no longer an explicit mechanism for deriving scores for each domain. Instead, pass/fail/refusal scores are assigned to all items. Item performance is then reinterpretated in relation to children s ages in terms of caution, delay, no opportunity, or normal or advanced performance. Two or more delays produce an abnormal overall test score while one delay and/or two or more cautions result in a questionable score. Restandardization was conducted exclusively in Colorado on 2096 children who were stratified by age, race, socioeconomic status, and residential area. Separate normative studies were conducted in Oklahoma, Tennessee, and Michigan. Some of these show high rates of questionable and abnormal scores (30% to 45%) but otherwise hold to expected patterns: abnormal results increased with children s ages and decreased the more educated the parents.7 Despite the apparent improvements, the Denver-Il was published without information about its validity or accuracy in identifying children with and without problems. Thus, there is little proof that the revised measure actually represents a viable improvement over its predecessor. The goal of the present study was to evaluate the accuracy of the Denver-I! by comparing its performance to standards for screening tests. These included sensitivity (the percentage of children with true problems who are correctly detected-approximateby 80% is preferable); specificity (the percentage of children without problems who are correctly detected-since there are many more children without problems, 90% is preferable); and positive predictive value (of the children who fail the screening test, the percentage who have true developmental problems on diagnostic testing-70% or about three of every four referrals).8 9 Subjects and Sites METHODS Five day-care centers were selected as sites to facilitate naturalistic sampling on the basis of race, age, and socioeconomic status. Subjects were 1 04 children, representing the majority of enrollees across the five centers. The majority (n = 72) attended one of three centers with state and federal tuition subsidies (Title XX) in which 73% of parents had incomes below poverty guidelines. Sixty percent of parents were unmarried and all were either employed or attended school. Parents levels of education approximated or PEDIATRICS Vol. 89 No. 6 June

2 exceeded national averages according to the 1990 US Census: 19% had not graduated from high school, 36% had high school diplomas; 25% had post-high school training, and 20% had at least 4 years of college. The majority of the children in the three subsidized centers were black (97%). The remaining 32 subjects, 94% of whom were white, attended one of three nonsubsidized day-care programs. Only 9% of their parents were unmarried and most had high levels of income and education: 3% had not graduated from high school and 80% had at least 4 years of college. Across all five centers, the subjects ages ranged from 3 to 72 months with a mean age of 39 months and a standard deviation of 17.1 months: 7% were 0 through I 1 months old; 1 1 %, 12 through 23 months; 21%, 24 through 35 months; 26%, 36 through 47 months; 18%, 48 through 59 months; and 17%, 60 through 72 months. There were an equal number of boys and girls. Procedures Each subject was administered the Denver-lI in strict accordance with the Denver-Il examiner s manual, by a licensed psychological examiner. A second examiner, blind to the results of the Denver- II, administered a widely recognized battery of tests which varied only according to the age of the child. The concurrent battery was administered within 1 week of the Denver-Il and was designed to detect the more common developmental disabilities including mental retardation, speech-language impairments, and learning disabilities. Multiple-converging measures were used to minimize any weaknesses within individual tests. The battery also employed multiple measurement methods (direct elicitation, observation, and parent report), which ensured the ecological validity of the results. The battery included (1) an individual measure of cognitive skills (either the Bayley Scales of Infant Development, Kaufman Assessment Battery for Children, or the Stanford-Binet Intelligence Scale, 4th edition); (2) the Vineland Adaptive Behavior Scale, which is a parent-report measure of socialization, communication, self-help, and motor development; and (3) the Kaufman Assessment Battery for Children Achievement Subtests (measures of language and academic skills for children 3#{189}years and older). Language skills (in children 2 through 6 years of age) were also assessed with the Fluharty Preschool Speech and Language Screening Test, #{176}a foursubtest measure of vocabulary, syntactic understanding, expressive syntax, and articulation. The Fluharty is widely used in schoolbased screening programs due to its high correlations (.897) with diagnostic speech-language measures such as the Peabody Picture Vocabulary Test and its accuracy in relation to criterion language measures (90% agreement).#{176}12 The Fluharty was standardized on 2147 children stratified by race/ethnicity, socioeconomic status, and geographic area of residence and has high levels of interrater and test-retest reliability (0.95 to I.0). The presence or absence of a developmental diagnosis was determined by having one of the examiners apply (usually federal and state criteria defining handicapping conditions for special education eligibility) to the results of the concurrent battery. Specifically, the following conditions were diagnosed: (1) mental retardation when IQ and adaptive behavior scores fell below 70 (Except that for children 30 months and younger, all of whom were administered the Bayley, mental retardation was diagnosed only if the Mental Developmental Index score was less than 50 and the adaptive behavior score less than 70. Stringent criteria were used for scoring the Bayley because of the antiquity of its norms and because of the high levels of stability in scores less than 50.) ; (2) functional developmental delay when IQ was below 70 (Bayley less than 50) but adaptive behavior score was 70 or above; (3) slow learning (children aged 31 months and older only) when IQ was between 70 and 79; (4) learning disabilities when achievement score was 85 or below and I standard deviation below IQ; (5) autism or related disorders if they met criteria listed in the Diagnostic and Statistical Manual ofmental Disorders, 3rd edition, revised; and/ or (6) language delays when children failed the three language subtests of the Fluharty or had standard scores below the 9th percentile on the language subtests of the Kaufman or Vineland. RESULTS Eighteen (1 7%) of the 1 04 children were found to have developmental diagnoses including mental retardation (2%), learning disabilities (1%), slow learning (6%), functional developmental delay (1 %), autism (1 %), and language delays (7%). The remainder did not meet diagnostic criteria and so were identified as nonhandicapped. The results of the Denver-Il were as follows: 40 (38%) children received normal scores; 27 (26%) had abnormal scores; 34 (33%) had questionable results; and 3 (3%) were deemed untestable because of refusabs or other oppositional behaviors. There are several different ways to intersect Denver-II results with diagnostic test data. The first is to group passing with questionable/untestable scores since children with the latter are not immediately referred for evaluations or interventions. Accordingly, as shown in Fig. 1, the Denver-I! identified 10 of the 1 8 children with handicaps. Thus, its sensitivity was 56%. Specificity was 80% since 69 of the 86 children without diagnoses passed the Denver-Il. The overall hit rate (total number of children correctly classified) was 76% and the positive predictive value was 37% because 1 0 of the 27 children who failed the Denver- II received a developmental diagnosis. There is better justification for grouping questionable/untestable with abnormal results because if children receive either a questionable or an untestable score twice across a 90-day interval, they are referred for diagnostic testing. Additionally, children are likely to have similar scores over brief time periods (as is evident from the Denver-II s high levels of test-retest reliability over a 7- to 10-day interval).7 As shown in Fig. 2, when grouping abnormal with questionable/ untestable scores, the Denver-Il identified 15 of the 1 8 children, including 71 % of the children with banguage delays, 83% of those with slow learning, and DENVER-Il PASS DEVELOPMENTAL DEVELOPMENTAL DIAGSES FAILr 10 Specificity 69/86 80% Sensitivity 10/18 56% Positive Predictive Value 10/27 37% Overall Hit-Rate 79/104 76% Fig 1. Accuracy of the Denver-Il (with questionables/untestables as passes). DENVER-li FAIL DIAGSES Specificity 37/86 = 43% Sensitivity 15/18 = 83% Positive Predictive Value 15/64 23% Overall Hit-Rate 52/104 = 50% Fig 2. Accuracy of the Denver-lI (with questionables/untestables as fails) DEVELOPMENTAL SCREENING

3 100% of the children with autism, learning disabilities, developmental delays, and mental retardation. Its overall sensitivity was 83%. Specificity was 43% since 37 of the 86 children without a diagnosis passed the Denver-Il. Overall, 50% of the children were correctly classified. Of the 64 children who failed the Denver-Il, 15 received a diagnosis such that the predictive value of a positive Denver was 23%. Given the abundance of abnormal and questionable Denver-Il results, we raised the hypothesis that the Denver-Il may have identified children who, while they did not appear to meet criteria for special education, were nevertheless experiencing subtle problems. If detected by screening such children could become the beneficiaries of increased monitoring and informal developmental stimulation. To test this hypothesis, a second set of criteria defining subtle developmental problems was applied to the diagnostic test results (eg, IQ or language, or achievement between the 9th and 1 6th percentile [standard scores between 80 and 84] or failure on 1 or more Fluharty subtests). Using these criteria, an additional 18 children appeared to be experiencing subtle developmental problems. Figure 3 shows the intersection of both subtle and diagnosable developmental problems with Denver-Il results (grouping questionable and untestable scores with abnormal scores). Sensitivity was found to be 75% since 27 of 36 children experiencing difficulties failed the Denver-I!. Specificity was 46% since 31 of 68 children without difficulties passed the Denver-Il. The overall hit rate was 56%. The positive predictive value was 42% since 27 of the 64 children with abnormal Denver-Il results had developmental difficulties. Because the Denver-Il identified as abnormal many children without any evidence of developmental problems, an effort was made to pinpoint possible sources of error. First, we tested the hypothesis that the age differences contributed to Denver-Il maccuracy. To do this, we grouped children by ages: 0 through 23 months, 24 through 47 months, and 48 through 72 months. Sensitivity, specificity, and posifive predictive value were computed for each group. Across all age levels sensitivity was relatively high, ranging from 73% in the 25- through 47-month group to 100% in the 0- through 24- and 48- through 72- month groups. Specificity remained fairly low: 31% for 48- through 72-month-olds, 44% for 0- through 23-month olds, and 53% for 24- through 47-montholds. Positive predictive value ranged from 18% to 3 1 %. Thus the Denver-Il appeared to overrefer across all ages. A second hypothesis was that the scoring criteria were too stringent (eg, a child can fail with delays on only two items or receive a questionable score on the basis of only two cautions or a single delay). Accordingly, we viewed whether children who had a greater number of delays and cautions were more likely to be true positives. This graphic analysis revealed the expected asymptotic decline in the overreferral rate with an increase in the numbers of failed items. However, there was a concomitant decrease in the Denver-II s detection of children with true problems, suggesting that the test s inaccuracies were only partially a function of scoring. A third hypothesis was that the content of the Denver-I! differed significantly from that of criterion measures. Since the criterion battery had only a bimited sample of gross motor skills and these did not enter into the diagnosis of handicapping conditions, we excluded the gross motor domain of the Denver- II and redefined as normal those children who failed the Denver-Il only on the basis of failed items in the gross motor section. This resulted in a slight drop in sensitivity (from 83% to 78%) and a slight increase in specificity (from 43% to 51%) and positive predictive value (from 23% to 25%). Overall, 14% of inaccurate Denver-Il results were the result of delays only in the gross motor domain. However, many more overreferrals (39%) were the result of failures in the fine motoradaptive and/or personal-social domains, which are areas of development measured by the concurrent battery. This suggests that much of the Denver-II s inaccuracies are due to inappropriate item content and placement in relation to children s ages. The remaining overreferrals (47%) resulted from delays in the language domain. However, the language domain also significantly discriminated delayed from nondebayed children. Referral decisions based on this solution, as presented in Fig 4, show that 67% of the children with diagnoses and 73% of the children without diagnoses could be identified by administering only the language domain and then applying the overall scoring criteria. A final alternative was to use failure of the language domain plus one or more other DENVER-Il SUBTLE AND SIGNIFICANT DIFFICULTIES TT TT1 : T-T---r] PASS 31 9 FAIL Specificity 31/68 = 46% Sensitivity 27/36 75% Positive Predictive Value 27/64 42% Overall Hit-Rate 58/104 = 56% DENVER-Il PASS FAIL DEVELOPMENTAL. - DIAGSES F_;i-. : Specificity 63/86 = 73% Sensitivity 12/18 = 67% Positive Predictive Value 12/35 34% Overall Hit-Rate 75/104 = 72% Fig 3. Accuracy of the Denver-Il (with questionables/untestables as fails compared to subtle and significant difficulties). Fig 4. Accuracy of the Denver-Il (presence or absence of delays in language domain). ARTICLES 1223

4 domains as scoring criteria. However, this resulted in high specificity, 87%, but limited sensitivity, 56%. DISCUSSION The results showed that most children with developmental problems, including problems that were somewhat subtle, received abnormal, questionable, or untestable scores on the Denver-Il. Thus, the measure appears to have sensitivity which approaches standards for screening tests. However, almost half the children without developmental problems also received suspect scores on the Denver-Il. This means that approximately three of every five children taking the test would be referred for evaluations, and fewer than one of the three referred would have true problems. Such inaccuracies are likely to waste precious diagnostic resources and result in needless parental anxiety and expense. Because of the exceedingly high overreferral rate, an effort was made to search for the sources of error within the test. Children who had delays or cautions solely in the fine motor-adaptive, personal-social, or gross motor domains were often those overreferred. Only the presence or absence of delays or cautions in the language domain discriminated children with and without problems. However, the use of this domain alone for making referral decisions would result in the underdetection of more than one third of the children with true developmental problems and a high overreferral rate. Several questions can be raised by the above findings, including whether it is fair to compare the Denver-Il to a broad-ranging test battery. Certainly, many studies of screening tests involve comparisons only with intelligence tests. Yet, screening tests measure multiple aspects of development including language, motor, self-help, and socialization. As a consequence, most current screening tests, including the Denver-Il, are designed to detect not just the 2% to 3% of children with mental retardation but also those with a range of mild and less-than-obvious developmental problems such as language impairments, devebopmental delay, learning disabilities, etc. Thus, it is not surprising that in this study, almost 20% of the children were found to have developmental problems. The use of a broad concurrent battery ensured that the strengths and weaknesses of the Denver-Il could be elucidated, and elucidated in relation to the more common developmental problems found in children. As a consequence, it can be seen that the Denver-Il represents a tremendous improvement over earlier editions in that it is sensitive to quite subtle developmental difficulties. Limitations in this study include a wide age range of subjects and a limited sample of children younger than 24 months of age. However, the consistency of the findings mitigates this weakness since the Denver- II had high sensitivity and limited specificity across all ages. The relative antiquity of some of the tests in the concurrent battery also raises concerns. However, all tests in the battery have been subjected to extensive research and are validated for children of the same ages and diagnoses as those in this study. 3 4 Any weaknesses in individual tests are further minimized by the study s use of multiple, converging diagnostic measures because no single test served as the sole criterion. Finally, the consistency of findings across varying tests and diagnoses supports the conclusions of this study. The Denver-II s weaknesses are additionally corroborated by observations from actual practice. The State of Tennessee s pilot efforts to implement Public Law involves a single-point-of-entry referral system. Referrals from those regions in Tennessee which rely heavily on the Denver-Il have included a large number who were not found to have developmental problems upon further assessment. Referrals appear to be more accurate from regions relying on other tests (Newlin-haus E, Coordinator, Early Intervention Hotbine, telephone communication, May 1991). The above observations and empirical findings suggest that additional development of the instrument is needed. The Denver-II s extensive restandardization, while commendable, represents only one aspect of recommended standards in test construction. 5 Validation in comparison with widely accepted diagnostic tests is essential since this is the only method for determining how well each item performs in relation to diagnostic measures or other performance criteria (eg, school success) and wheth#{128}ritems actually measure meaningful aspects of child development. In the case of screening tests, validation should also include assessment of accuracy in decision making, which aids in the determination of a measure s final scoring criteria. 6 The fact that these crucial aspects of test construction were omitted from the development of Denver-Il may explain why the test lacks accuracy. Because of copyright restrictions, the modifications suggested by the present study can be made only by the authors of the Denver-Il, who are urged to conduct a large-scale validity study of their own. Based on the present results, expected changes should indude the following: eliminating or replacing items that fail to discriminate performance on diagnostic tests; revising scoring criteria, perhaps by weighting performance in the language domain; and clarifying referral criteria in relation to performance across domains. The test protocol will need to be revised to reflect these changes as will the examiner s manual. In the interim, those involved in developmental screening are encouraged to use tests that have high levels of sensitivity and specificity. For example, the Battelle Developmental Inventory Screening Test 7 appears well-supported by a growing number of validity studies.8 8 Perhaps more functional for pediatric practice are the screening level measures from the Minnesota Child Development Inventory since these rely on parental report and are thus useful when children are refractory, ill, of fearful. 92 Of six validity studies on the Minnesota, five were positive and in the sixth, the authors failed to notice that all delayed subjects failed one or more subtests of the Minnesota and thus were correctly identified.2228 Although all screening tests produce some error, the decisions made on the basis of screening are consequential and thus dictate selection of instruments with the greatest levels of both sensitivity and specificity DEVELOPMENTAL SCREENING

5 Accurate screening contributes to parental well-being, helps distribute limited diagnostic services and health care dollars in the most parsimonious manner, and helps ensure that those children who need intervention are identified as early as possible. REFERENCES I. Borowitz KC, Glascoe FP. Sensitivity of the Denver Developmental Screening lest in speech and language screening. Pediatrics. 1986;78: Meisels SJ, Margolis LH. Is the early and periodic screening, diagnosis, and treatment program effective with developmentally disabled children? Pediatrics. 1988;81: Meisels SJ. Can developmental screening tests identify children who are developmentally at risk? Pediatrics. 1989;83: Applebaum A. Validity of the revised Denver Developmental Screening lest for referred and non-referred samples. Psychol Rep. 1978;43: Sciarillo WG, Brown MM, Robinson NM, Bennett FC, Sells CJ. Effectiveness of the Denver Developmental Screening Test with biologically vulnerable infants. I Dev Behav Pediatr. 1986;7: Frankenburg WK, Dodds J, Archer P, et al. Denver-li: Screening Manual. Denver, CO: Denver Developmental Materials; Frankenburg WK, Dodds J, Archer P, et al. Denver-li: Technical Manual. Denver, CO: Denver Developmental Materials; Wolery M. Child find and screening issues. In: Bailey DB, Wolery M, eds. Assessing Infants and Preschoolers With Handicaps. Columbus, OH: Merrill Publishing Company; 1989; Glascoe FP, Martin ED, Humphrey S. A comparative review of developmental screening tests. Pediatrics. 1990;86: Fluharty NB. Fluharty Preschool Speech and Language Screening Test. Allen, TX: DLM-Ieaching Resources; Schetz KF. Comparison of the Compton Speech and Language Screening Evaluation and Fluharty Preschool Speech and Language Screening lest. Lang Speech Hear Serv Sch. 1985;16: Illerbrun D, Haines L, Greenough P. Language identification screening test for kindergarten: a comparison with four screening and three diagnostic language tests. Lang Speech Hear Serv Sch. 1985;16: SattlerJM. Assessment of Children. 3rd ed. San Diego, CA: Jerome Sattler; Kline RB. Is the fourth edition Stanford-Binet a four-factor test? Confirmatory factor analyses of alternative models for ages 2 through Psychoeduc Assess. 1989;7: American Psychological Association. Standards for Educational and Psychological Tests. Washington, DC: American Psychological Association; Frankenburg WK. Selection of diseases and tests in pediatric screening. Pediatrics. 1974;54: Newborg J, Stock JR. Wnek L, Guidubaldi J, Svinicki J. Battelle Developmental Inventory Screening Test. Allen, IX: DLM-leaching Resources; Wossum D. The validity of the Battelle Developmental Inventory Screening Test for early detection of developmental disorders. Diss Abstr Int B Sci Eng. 1991;52: Ireton H, Thwing E. Minnesota Infant Development Inventory. Minneapohs, MN: Behavior Science Systems; lreton H, Thwing E. Early Childhood Development Inventory. Minneapolis, MN: Behavior Science Systems; Ireton H, Thwing E. Preschool Development Inventory. Minneapolis, MN: Behavior Science Systems; Guerin D, Gottfried AW. Minnesota child development inventories: predictors of intelligence, achievement and adaptability. I Pediatr Psychol. 1987;12: Chaffee CA, Cunningham CE, Secord-Gilber M, Elbard H, Richards J. Screening effectiveness of the Minnesota Child Development Inventory Expressive and Receptive Language Scales: sensitivity, specificity, and predictive value. I Consult Clin Psychol. 1990;2: Creighton DE, Suave RS. The Minnesota Infant Development Inventory in the developmental screening of high-risk infants at eight months. Can I Behav Sci. 1988;20: Saylor CF. Brandt BJ. The Minnesota Child Development Inventory: a valid maternal-report form for assessing development in infancy. I Dev Behav Pediatr. 1986;7: I 26. Sturner RA, Funk 5G. Thomas PD, Green JA. An adaptation of the Minnesota Child Development Inventory for preschool developmental screening. I Pediatr Psychol. 1982;7: Kopparthi R, McDermott C, Sheftel D. The Minnesota Child Development Inventory: validity and reliability for assessing development in infancy. I Dev Behav Pediatr. 1991;12: Glascoe FP. Letter to the editor re: Kopparthi R, McDermott C, Sheftel D. The Minnesota Child Development Inventory: validity and reliability for assessing development in infancy. / Dev Behav Pediatr. In press. THE HIGH-TECH EFFECT...if a doctor carefully questions a patient for 1 0 to 20 minutes, thoroughly examines that patient for another 10 to 20 minutes, and then counsels that patient for 5 to 10 minutes, he will be lucky to get $60 from the insurance company. Yet when a radiologist reads an M.R.I. scan for approximately 15 to 20 minutes, he will be unlucky if the insurance company pays him less than $200. The formula becomes: if it s a high-tech test, it s less time and more money. If units of time are the measuring units, the discrepancy can be 1 0 times greater! Vigman MP. High-tech medicine pays physicians best. The New York Times. June 26, Submitted by Student ARTICLES 1225

6 Accuracy of the Denver-II in Developmental Screening Frances Page Glascoe, Karen E. Byrne, Linda G. Ashford, Katherine L. Johnson, Bernard Chang and Bryan Strickland Pediatrics 1992;89;1221 Updated Information & Services Permissions & Licensing Reprints including high resolution figures, can be found at: Information about reproducing this article in parts (figures, tables) or in its entirety can be found online at: Information about ordering reprints can be found online:

7 Accuracy of the Denver-II in Developmental Screening Frances Page Glascoe, Karen E. Byrne, Linda G. Ashford, Katherine L. Johnson, Bernard Chang and Bryan Strickland Pediatrics 1992;89;1221 The online version of this article, along with updated information and services, is located on the World Wide Web at: Pediatrics is the official journal of the American Academy of Pediatrics. A monthly publication, it has been published continuously since Pediatrics is owned, published, and trademarked by the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elk Grove Village, Illinois, Copyright 1992 by the American Academy of Pediatrics. All rights reserved. Print ISSN: