[Helen Carey]

Criterion referenced tests are designed to measure patient performance against predetermined criteria, instead of normative data distributed across a bell curve. These tests are often a series of skills or behaviors measured against a set criteria for performance. Instead of scaled or composite scores, these tests often have cut-off scores to compare individual performance against a description of performance and not against group performance (normative data). Unfortunately, criterion referenced tests don’t all follow the same structure/format. For example, criterion tests might group expected skills for particular age ranges in separate sections. A child who can perform all skills in that age section is described as having mastery of those type of skills (such as gross motor) at least at that age cut off. Another example is the GMFM, where items are organized from easiest to hardest in each dimension. It doesn’t have cut offs based on any criteria, other than the individual child’s performance (each time the test is administered, the child basically creates their own cut off scores). Some examples of criterion referenced tests are the Ages & Stages Questionnaire; the Assessment, Evaluation, and Programming System for Infants and Children (AEPS); the WEEFIM; the PEDI (scaled scores only); the GMFM; School Function Assessment (SFA); TUG test; and Pediatric Balance Scale.

Evaluative tests are designed to measure change over time, therefore, we need to use tests that are sensitive to change and allow comparison with subsequent test administrations. Criterion-referenced tests allow for this type of comparison over time. For each administration, the same items are administered, therefore, the criteria for performance doesn’t change. In a norm-referenced test, the criteria changes with subsequent administrations when the child’s raw score is compared against a new set of standards (or a new age bracket of normative data).

Normative data is derived from testing hundreds of infants/children (depending on the type and age range of test) and the plotting their scores on a bell curve (“normal distribution”). Because children can have variable performance, the bell curve captures this variance and plots it as a distribution under the curve.

[Helen Carey]

For the Bayley-III, 1 standard deviation (SD) below the mean and above the mean correlate with the following: Scaled score range 7-13 (3 points below and above the mean of 10) and composite score range 85-115 (15 points below and above the mean of 100). 1 standard deviation also correlates with percentile rank equivalents of 16 (1 SD below the mean) and 84 (1 SD above the mean). There are pictures of bell curves online that display all of these score correlations, as they are standard psychometrics, not unique to the Bayley. The picture might help you visualize the correlations.

Relative to the practice test question, a scaled score of 9 is well within 1 SD of the mean (although below the mean of 10) so is classified as “average.” The Bayley-III manual classifies scores of 8 or greater as “average” as they are within 1 SD of the mean. Scaled scores of 6 and 7 are low average (or often interpreted clinically as borderline) while scores below 6 are “below average.”

[Helen Carey]

Hi Krina,
The Bayley is a discriminative measure as each administration results in scores based on comparisons to normative values (the normal sample distributed on a bell curve). Therefore, you are “discriminating” the child’s performance from that of typically developing peers.

Evaluative measures are designed to “evaluate” a child’s performance over time (or from at least 1 administration to another). These scores are not plotted against normative data on a bell curve. The initial administration of the test is often a baseline assessment, which serves as a comparison to future administrations. Therefore, you can evaluate change over time. Values such as the MDC and MCID help us interpret the meaningfulness of the child’s change. Examples of evaluative measures are the GMFM and the TUG test.

The Bayley is not designed to evaluate a child’s performance over time (comparison of scores from multiple administrations) as each administration results in scores derived from a new table of normative data. Therefore, at each administration the child is compared to a new normative sample, not themselves. The Bayley is not capable of generating change scores like the GMFM-66.

We hope this explanation is helpful!
Helen and Jessica

[Author]

Posts