Background. In 2001, the American Academy of Pediatrics (AAP) adopted a policy that all infants and young children should be screened for developmental delays at regular intervals. The policy statement promoted the use of valid reliable instruments. It is unknown, however, what proportion of pediatricians follow this recommendation and whether such a practice is associated with improved identification of children with developmental difficulties.
Objectives. To describe the use of developmental screening tests among board-certified pediatricians practicing general pediatrics and to determine the association between standardized screening and the self-reported identification of children with developmental difficulties.
Methods. We mailed a survey to a random sample of AAP members. We used multivariate logistic/linearregression analyses to determine the association between standardized screening and the self-reported identification of children with developmental disabilities.
Results. Of the 1617 surveys mailed, 894 were returned, for a response rate of 55%. Of the respondents, 646 practiced general pediatrics and were included in the analysis. Seventy-one percent of those pediatricians indicated that they almost always used clinical assessment without an accompanying screening instrument to identify children with developmental delays. Only 23% indicated that they used a standardized screening instrument. The most commonly used instrument was the Denver II. Logistic regression modeling demonstrated odds ratios between 1.71 and 1.90 for a >10% rate of identification of developmental problems among patients of pediatricians reporting standardized screening. Each adjusted odds ratio bordered on statistical significance. Linear-regression models estimating the difference in mean proportions of children identified with developmental problems across screening groups failed to show a statistically or clinically significant difference in physician-reported identification rates.
Conclusions. Our findings indicate that, despite the AAP policy and national efforts to improve developmental screening in the primary care setting, few pediatricians use effective means to screen their patients for developmental problems. It is uncertain whether standardized screening, as it is practiced currently, is associated with an increase in the self-reported identification of children with developmental disabilities.
Children with developmental difficulties constitute between 12% and 16% of the general pediatric population.1 Early diagnosis of such children is important, because evidence has demonstrated the effectiveness of early intervention for children with many developmental conditions.2–6 Because 95% of children from birth through 3 years of age report a regular source of health care7 and because pediatricians have frequent contact with infants and young children, many think that primary care pediatricians are uniquely suited for the detection of young children with developmental difficulties.8,9
Although only one half of families report that their children have ever received a developmental assessment performed by their doctors,10 virtually all general pediatricians report assessing developmental milestones as part of routine well-child care.11 The methods by which they do so vary, however. Previous studies indicated that the most common developmental screening technique used in the primary care setting is informal clinical assessment1 and that few pediatricians use standardized developmental screening tests routinely during health supervision visits.12,13 Unfortunately, clinical assessment alone detects <30% of children with developmental disabilities12 and, at the present time, only 20% to 30% of children with disabilities are identified before school entrance.14 In contrast, standardized developmental screening instruments are reported to have sensitivities and specificities of 70% to 90%.15–18
These factors led the American Academy of Pediatrics (AAP) to publish a policy statement in 2001, calling for universal developmental screening of infants and young children as part of routine well-child care. The statement emphasized the use of standardized screening tools that are practical and easy to use in the office setting.19 Pediatrician practices in the area of developmental screening, however, have not been reexamined systematically since the release of this statement.
The main purpose of this article is to describe the current use of developmental screening tests among a nationally representative sample of board-certified pediatricians. We also examine the association between standardized screening and the self-reported identification of children with developmental disabilities.
Context and Sample
This survey represented the 53rd in the series of AAP Periodic Surveys of Fellows conducted by the Division of Health Policy Research of the AAP. Each periodic survey is mailed to a unique random sample of US AAP members. For this survey, 1617 pediatricians were selected from the 49530 US AAP members who were active at the time of the study. An estimated 80% of US board-certified pediatricians are members of the AAP. This particular survey was part of a collaborative project of the AAP, the Federal Maternal and Child Health Bureau, the Department of Education Office of Special Education Programs, and the Office of Special Education Programs-funded ChildFind Project to obtain national data identifying barriers to pediatricians' participation in early intervention programs and to explore ways to address those barriers.
The 8-page, self-administered survey was developed by an AAP-appointed work group of experts in early intervention from around the United States. The work group included general and developmental pediatricians, educators, physical and occupational therapists, psychologists, and parents. For the purpose of this survey, developmental delays or problems included motor, language, cognitive, behavioral, or emotional issues. The survey was approved by the AAP institutional review board.
The survey questioned respondents about the frequency with which they use a variety of formal and informal screening techniques to identify 0- to 35-month-old patients with developmental problems. Because the AAP policy statement promotes universal screening with a standardized instrument,19 we considered pediatricians to practice standardized screening only if they responded “always” or “almost always” to the question, “How often do you or your staff use the following methods or tools to identify children birth through 35 months of age at risk for developmental delay or problems?” (in relation to the Bayley neurodevelopmental screen, Denver II, Ages and Stages Questionnaire, Parents' Evaluation of Developmental Status, or an unspecified “other” instrument).
Respondents were also asked, “What percent of your current patients <36 months old have been identified with a possible developmental problem through assessments/screens performed in your office?” We analyzed this response both as a continuous variable and as a dichotomous variable, with a 10% cutoff value. We chose a 10% cutoff value because it represents a rounded conservative interpretation of the AAP's published estimate that 12% to 16% of children of this age have a developmental problem.
We used a 5-point Likert scale to assess barriers to performing developmental screening with young children. We scored Likert items positively if respondents “agreed” or “strongly agreed” with statements offered as possible barriers.
The demographic portion of the instrument included a series of questions validated by the AAP Department of Research.20 Reported Medicaid and managed care participation was dichotomized relative to the median value of each continuous variable (for Medicaid: 30% participation; for managed care: 60% participation).
The survey instrument was mailed to each potential respondent between May and September 2002. Each questionnaire was accompanied by an introductory letter from the Executive Director of the AAP and a business-reply return envelope. Potential respondents received up to 6 mailings.
Only pediatricians who provided health supervision were eligible to respond to survey items on developmental screening and referrals. We also restricted our analysis to pediatricians who reported spending ≥10% of their time in general pediatrics. From this pool of generalists, we excluded respondents who reported spending any time in a development-oriented specialty, including developmental or behavioral pediatrics, neurodevelopment/disabilities, neonatology or perinatology, genetics, and neurology.
For the study-eligible population, we calculated 4 regression models to describe the association between pediatricians' report of standardized screening and the self-reported identification of patients 0 to 3 years of age with a developmental problem. We used logistic regression to model the association between standardized screening and the identification of >10% of children with a developmental problem. We used linear regression to model the difference in means of reported identification rates between those reporting standardized screening and those not. For both the linear and logistic models, we restricted the analyses to pediatricians reporting an identification rate within 2 SDs of the total sample mean (≤32%). We imposed this last restriction because we thought it likely that responses higher than this would reflect either a misinterpretation of the survey item or an atypical practice setting. The regression models were as follows.
The first model was simple logistic/linear regression, showing the unadjusted association between standardized screening and physician-reported identification of children with developmental problems. The second model was multivariate logistic/linear regression, with the addition of only Medicaid participation to the simple model. We added Medicaid participation individually because of its empiric association with the report of both standardized screening and identification of children with developmental problems.
The third model was theoretically derived multivariate logistic/linear regression, in which covariates were included on the basis of their theoretical relevance as potential confounders of the relationship between standardized developmental screening and physician-reported identification of children with developmental problems. We chose the following covariates to include in this model: the gender and age of the pediatrician, managed care and Medicaid participation, the proportion of time spent in general pediatrics, whether the pediatrician practiced in a medical school or university setting, and the proportion of patients 0 to 3 years of age seen in the respondent's practice. The fourth model was backward stepwise logistic/linear regression, in which the least significant variables (age, managed care participation, proportion of time in general pediatrics, and proportion of patients 0–3 years of age) were removed sequentially from the full theoretical model until each remaining variable had a P value of ≤.2.
We chose to present 4 different models because there exists no previous research or rationale to suggest the validity of one over any of the others. In the absence of a clearly superior model, we considered it important to convey the stability of findings across a variety of acceptable models.
Of the 1617 surveys mailed, we received 894, for a total response rate of 55%. Respondents and nonrespondents were similar with respect to age, gender, AAP membership status, and geographic location (data not shown). The characteristics of the respondents in this survey were also similar to the known characteristics of the members of the AAP and to those of respondents to other Periodic Surveys conducted around that time.21,22
Table 1 represents a description of all respondents included in the analysis (n = 646). Of the 248 respondents who were removed from the analysis, 222 were excluded for practicing <10% of the time in general pediatrics and 26 for practicing a development-oriented specialty in addition to general pediatrics. Eleven percent of those included in the analysis practiced in a medical school or university setting; 43% were in practices of ≥2 pediatricians. On average, pediatricians included in the analysis reported spending 88% of their time in general practice and reported that one half of their patient base was <3 years of age.
Seventy-one percent of respondents reported always or almost always relying on nonstandardized methods to detect developmental problems among their 0- to 3-year-old patients (Table 2). Thirty-three percent reported always or almost always screening with a combination of nonstandardized methods and standardized tests, and 37% reported using a nonstandardized, office-generated checklist or similar method, typically completed by clinic staff members.
A minority of respondents (23% in total) reported always or almost always using a standardized screening instrument. Of all such instruments, the Denver II was the most commonly used. Only 41 pediatricians responded to the survey item concerning the use of other, unspecified, screening instruments; of those respondents, 49% reported using ≥1 such instrument. (It should be noted that multiple responses were possible; therefore, proportions do not add up to 100%.)
Characteristics of Respondents Using Standardized Tests Versus Nonstandardized Methods
Respondents using standardized tests were more apt to have a high proportion of Medicaid clients (60% vs 48%; P = .03) and were more likely to report a >10% rate of identification of developmental problems among their 0- to 3-year-old patients (26% vs 15%; P = .01) (Table 3). However, there were no differences in reported frequencies of children identified with developmental problems between those using standardized screening and those not (7.8% vs 6.8%; P = .07).
The results of multivariate regression models estimating the association between standardized screening and reported identification of developmental problems among children <3 years of age are shown in Table 4. Logistic regression models demonstrated odds ratios between 1.71 and 1.90 for a >10% reported rate of identification of developmental problems among pediatricians reporting standardized screening. The adjusted odds ratios did not change significantly across models, and each adjusted odds ratio bordered on statistical significance. Linear-regression models estimating the difference in the mean proportions of children reported to have been identified with developmental problems across screening groups failed to show a statistically or clinically significant difference; however, all regression coefficients demonstrated a trend toward higher reported identification rates among pediatricians reporting standardized screening.
Barriers to Screening
Table 5 shows the most widely reported barriers to standardized developmental screening in the office. Barriers included a lack of time, a lack of available office staff, and inadequate reimbursement.
Our study indicates that a majority (71%) of general pediatricians reported using clinical assessments or other nonstandardized methods to monitor the early development of their patients. Only 23% reported consistently using standardized screening; among those respondents, the Denver II was the most widely used instrument. Although we showed a trend toward greater self-reported identification of children with developmental problems among those using standardized screening instruments, our data do not support a statistically significant association between standardized screening, as it is currently practiced, and self-reported detection of developmental problems.
Our results are consistent with previous studies, which showed that 15% to 40% of pediatricians reported using standardized screening12,13 and which suggested that the 2001 AAP policy statement has not affected practice widely. Furthermore, our inability to demonstrate an association between standardized screening and improved self-reported detection of developmental delays raises the questions of whether the performance characteristics of these instruments, particularly the Denver II, are maintained in the real world of general pediatric practice and whether such instruments are being used properly. Pediatricians cite time limitations, lack of staff, and inadequate reimbursement as the main barriers to standardized developmental screening.
A recent report by Sices et al23 suggested that, in the absence of standardized screening, both pediatricians and family physicians are inconsistent regarding their referral patterns for children with possible developmental problems and tend not to consider important risk factors (such as parental concern) when deciding whether to refer patients. In the context of recent research on the importance of early brain development for optimal social and cognitive development24 and evidence that early intervention programs can improve outcomes for patients with developmental disabilities,2,3 improving, eliminating the variability in, and providing rigorous effectiveness data for developmental screening practices in the primary care setting are important.
Glascoe and Macias25 suggested several ways to incorporate developmental and behavioral screening into pediatric practice, including dissemination of evidence for using parental reports as a screening tool, links to information about local services, and information on coding and billing practices for adequate reimbursement for developmental screening. In addition, Halfon et al26 argued for systems-level changes that would allow enhanced primary care-based developmental services and Zuckerman and Halfon27 for policy-level efforts aimed at defining medical necessity and requiring reimbursement for these services.
Our study was limited by a number of factors. First, as with many surveys, social desirability biases might have compelled respondents to overestimate their attitudes and practices. Although we addressed this by defining positive responses conservatively, our cutoff values for certain variables are admittedly arbitrary. In addition, although our 55% response rate was consistent with normative values for survey research28,29 and respondents appeared representative of the members of the AAP, our findings may not be generalizable to all primary care pediatricians. The association between a higher self-reported rate of identification of developmental problems and screening practices might be confounded by nonmeasurable characteristics, such as respondents' personal interest in developmental issues or a different prevalence of developmental problems among patients of respondents who use standardized screening measures. Finally, it is not known how accurate pediatricians are at estimating the percentages of children in their practices whom they have identified as having developmental problems.
The results of this study should not be construed to mean that standardized, validated, screening tools fail to identify children with developmental problems. The performance characteristics of such tools demonstrate their accuracy to be well above that of informal methods,15–18 and 1 recent study showed that, when providers switched to standardized instruments, detection rates increased significantly.30 Our findings do, however, raise the concerns that systems of care that foster the proper use of adequate detection methods in the primary care setting continue to be elusive and that we may be missing an important window of opportunity to identify children's developmental problems and to intervene to alter their developmental trajectories favorably.
This study was supported by the US Department of Education Office of Special Education Programs (grant H02 MC00073) and the National Center of Medical Home Initiatives for Children With Special Needs. Additional support for Dr Silverstein was provided by the Weaver Family Foundation.
We thank all members of the AAP-appointed work group for this project, including Gilbert Buchanan, MD; Angela Capone; Tom Castonguay; Molly Cole; Glinda Foster Hill; Barbara Jackson; Amy Lacroix, MD; Gary Lerner, MD; Debra Nelson; Mike, Ann, and Lisa Moody; Carlos Quezada-Gomez; Cordelia Robinson; Patti Rosquist, MD; Beppie Shapiro; and Dennis Vickers, MD. We thank Kari Hironaka, MD, MPH, Frederick P. Rivara, MD, MPH, and Lynn Olson for thoughtful review of the manuscript and Thomas Koepsell, MD, MPH, Howard Bauchner, MD, and John Cook, PhD, for methodologic expertise.
- Accepted November 11, 2004.
- Reprint requests to (M.S.) Boston Medical Center, Maternity Building, 4th Floor, 91 East Concord St, Boston, MA 02118. E-mail:
The views expressed are those of the authors and not necessarily those of the American Academy of Pediatrics.
Dr Glascoe developed and has a financial interest in the Parent’s Evaluation of Developmental Status developmental screening tool. This instrument is mentioned in the article along with a variety of other instruments. It is by no means a focus of the article.
- ↵Glascoe FP. Early detection of developmental and behavioral problems. Pediatr Rev.2000;21 :272– 279
- ↵Guralnick MJ. The Effectiveness of Early Intervention. Baltimore, MD: PF Brooks Publishing; 1997
- ↵Zuckerman B, Parker S. New models of pediatric care. In: Halfon N, Schuster M, eds. Child Rearing in America. New York, NY: Cambridge University Press; 2002
- ↵Halfon N, Regalado M, Sareen H, et al. Assessing development in the pediatric office. Pediatrics.2004;113(6 suppl) :1926– 1933
- ↵Silverstein M, Grossman DC, Koepsell TD, Rivara FP. Pediatricians' reported practices regarding early education and Head Start referral. Pediatrics.2003;111 :1351– 1357
- ↵Glascoe FP. Parents' concerns about children's development: prescreening technique or screening test? Pediatrics.1997;99 :522– 528
- Glascoe FP, Byrne KE, Ashford LG, Johnson KL, Chang B, Strickland B. Accuracy of the Denver-II in developmental screening. Pediatrics.1992;89 :1221– 1225
- ↵Squires J, Bricker D, Potter L. Revision of a parent-completed development screening tool: Ages and Stages Questionnaires. J Pediatr Psychol.1997;22 :313– 328
- ↵American Academy of Pediatrics, Committee on Children with Disabilities. Developmental surveillance and screening of infants and young children. Pediatrics.2001;108 :192– 195
- ↵Brotherton SE, Tang SfS, O'Connor KG. Trends in practice characteristics: analyses of 19 periodic surveys (1987–1992) of Fellows of the American Academy of Pediatrics. Pediatrics.1997;100 :8– 18
- ↵Gupta VB, O'Connor KG, Quezada-Gomez C. Care coordination services in pediatric practices. Pediatrics.2004;113(5 suppl) :1517– 1521
- ↵Kline MW, O'Connor KG. Disparity between pediatricians' knowledge and practices regarding perinatal human immunodeficiency virus counseling and testing. Pediatrics.2003;112(5) . Available at: www.pediatrics.org/cgi/content/full/112/5/e367
- ↵Sices L, Feudtner C, McLaughlin J, Drotar D, Williams M. How do primary care physicians manage children with possible developmental delays? A national survey with an experimental design. Pediatrics.2004;113 :274– 282
- ↵Shonkoff CJ, Phillips DA. From Neurons to Neighborhoods: The Science of Early Childhood Development. Washington, DC: National Academy Press; 2000
- ↵Glascoe FP, Macias MM. How can you implement the AAP's new policy on developmental and behavioral screening? Contemp Pediatr.2003;4 :85– 104
- ↵Halfon N, Regalado M, McLearn KA, Kuo AA, Wright K. Building a Bridge From Birth to School: Improving Developmental and Behavioral Health Services for Young Children. New York, NY: Commonwealth Fund; 2003
- ↵Zuckerman B, Halfon N. School readiness: an idea whose time has arrived. Pediatrics.2003;111 :1433– 1436
- ↵Pelletier H, Abrams M. The North Carolina ABCD Project: A New Approach for Providing Developmental Services in Primary Care Practice. Portland, ME: National Academy for State Health Policy; 2002
- Copyright © 2005 by the American Academy of Pediatrics