OBJECTIVE: We studied the accuracy of the Parents’ Evaluation of Developmental Status (PEDS®) at school entry in predicting academic, language, and health-related quality of life (HRQoL) outcomes 2 years later.

DESIGN/METHOD: Prospective population study in elementary schools in Melbourne, Australia. Base Population. A total of 1591 school entry children who participated in a separate cross-sectional study. Cases. One hundred seventy-three children randomly selected from those with significant parental developmental concerns. Controls. One hundred twenty-nine children without significant parental development concerns.

Predictors (1997). Parents completed all 10 PEDS® items; teachers completed 5 PEDS® items. Outcomes (1999). The PEDS® and the Child Health Questionnaire (parents) and the Comprehensive Inventory of Basic Skills-Revised and the Renfrew Action Picture Test of language (children). Odds ratios (ORs) were calculated for low outcome scores, defined as >1.0 standard deviation below the mean or <16th centile.

RESULTS: At outcome, case parents reported more “significant” concerns on the PEDS® than did control parents (65% vs 26%, OR 5.3), but mean language and academic scores were only slightly lower for case children. Parent-reported self-help and school skills concerns predicted low language (ORs 2.1-2.8) and academic (ORs 1.3-6.6) scores. Teacher concerns about early school skills predicted low academic scores (ORs 3.7-4.7). However, sensitivity and specificity values were modest. Baseline developmental concerns predicted poor scores on a number of domains of HRQoL 2 years later.

CONCLUSIONS: Although individual developmental concerns at school entry variably predict later academic and language scores, sensitivity and specificity values would not support use of the PEDS® as a stand-alone screen to detect later problems.

Comments: In the study, parents of five-year-olds were administered PEDS1 and their children’s academic and language outcomes viewed two years later. Fifty-four percent of the sample consisted of children whose parents held concerns associated with high rates of developmental disabilities.2 Why did the authors not track what happened in the time elapsed? Other than in-grade retention, the study does not account for such interventions as homework assistance, tutoring, remedial programs, summer school, or even special education placement. Given the strong association between parents’ discussion of concerns and children’s receipt of interventions, it seems likely that many parents would have taken actions leading to improvements in their children’s outcomes.3 If so, the long-term predictive power of PEDS® would have been masked. It would have been particularly helpful had the authors had assessed performance on concurrent measures at age 5 and compared this to performance at age 7 in order to account for positive trajectories. In any case, failure to account for interventions during the two-year interval, is a serious limitation in the study and should be carefully addressed.

For a more detailed discussion of issues to consider in predictive validity studies, please see this article written by several screening test authors and research:

The Thorny Nature of Predictive Validity Studies on Screening Tests for Developmental-Behavioral Problems PEDIATRICS Vol. 122 No. 4 October 2008, pp. 866-868 (doi:10.1542/peds.2007-3142)

Kevin Marks, MDa, Frances Page Glascoe, PhDb, Glen P. Aylward, PhDc, Michael I. Shevell, MDd, Paul H. Lipkin, MDe,f, Jane K. Squires, PhD

The Thorny Nature of Predictive Validity Studies on Screening Tests for Developmental-Behavioral Problems

Related Files:
Predictive Validity