Training the indigenous nonprofessional: the screening technician

Combining Two Developmental Screening Tests to Improve Predictive Accuracy

OBJECTIVE

Single developmental screening tests are insufficient for supporting clinical decisions, showing less than 50% of those referred are actually delayed. Combining two tests can address this problem. We used likelihood ratios (LR) to evaluate the predictive value (PV) of combining the Ages and Stages Questionnaires (ASQ) and Parents Evaluation of Developmental Status (PEDS) screening tests.

METHODS

We used data from Limbos and Joyce that included information on the ASQ, PEDS, and an age-appropriate diagnostic assessment of delay on 328 children (ages 12-60 months) drawn from primary care. LRs were calculated for each of the available ASQ/PEDS combinations then rank ordered. PV was examined using different LRs to define referral and nonreferral thresholds.

RESULTS

Setting the referral threshold at LR 2.0 for combined scores decreased the number of false positives from 66 to 28, increased false negatives from 6 to 9 with positive predictive value (PPV) improving significantly from 30% to 47% (P < .05). Sensitivity dropped from 82% to 74% and specificity improved from 72% to 90%. When the LR referral threshold increased, PV improved but an inconclusive group emerged that required more information for decision-making.

CONCLUSIONS

Combining the ASQ and PEDS significantly improved predictive accuracy of a referral in a single threshold analysis. Predictive value improved further as requirements for referral increased and an inconclusive group needing further study was recognized. Systematic research is needed to identify what readily available information will assist decision making in this group.

Validity of the Revised Denver Developmental Screening Test for referred and nonreferred samples

The Revised Denver Developmental Screening Test was compared with the Bayley Scales of Infant Development to investigate the validity of the Denver scale for children under 30 mo. of age. Two groups of infants were studied, referred and nonreferred samples. Correlations of .25 to .52 were obtained between the two measures on the basis of developmental quotients which are significantly lower than those previously reported as mental age estimates. Generally the Denver scale appears to be a conservative test in terms of its classifications, and there is a marked underselection by the Denver test of children who receive abnormal scores on the Bayley scales. There is, however, congruence between the Denver Developmental Quotient and the Bayley Mental Scale score for children of lower ability. While two Denver mental age estimates (personal-social and gross motor) appear congruent with the Bayley? Mental Scale age equivalent, for the most part the Denver test yields significant underestimates of mental age. Over-all the data offer only limited validational support for the use of the Denver test with infants under 2 1/2 yr. of age.