In IRT, each response pattern results in a different level of functioning (
θ) and an associated reliability, expressed as the standard error of theta (SE(
θ)). A SE(
θ) of 0.32 or lower was considered a reliable measurement, which corresponds to a reliability of 0.90 or higher. To investigate the reliability of the Peer Relationships item bank and short form,
θ estimates and SE(
θ) were calculated using the Expected A Posteriori (EAP) estimator. Post hoc CAT simulations were performed on the respondent data with the R-package “catR (v3.16)” [
30] using maximum posterior weighted information (MPWI) selection criterion and EAP estimator [
31] to assess how a CAT would perform when applying the Dutch model parameters. The starting item was the item that offered most information at the mean of the study sample (
θ = 0). The stopping rules for the CAT were a maximum of eight items administered (which is equal to the length of the short form) or a SE(
θ) < 0.32 [
32]. To compare the reliability of the full item bank, short form, and CAT with the PedsQL social functioning scale, a GRM model was also fit to the PedsQL data and
θ estimates and SE(
θ) were calculated and presented in a reliability plot. In a reliability plot, each line represents the standard errors of measurement across
θ or T-score of one measure. A lower line is indicative of a higher reliability. Plotted dots are individual estimated thetas or T-scores and their associated standard errors of measurement resulting from post hoc CAT simulations. The current PROMIS convention is to use the U.S. parameters model for calculating
T-scores, unless significant differences are found between country-specific model parameters and the U.S. parameters. Therefore, the reliability of measurements were also calculated using the U.S. parameters (provided by HealthMeasures) and plotted in a reliability plot and included the T
-score distribution of the Dutch population as histogram. In addition, efficiency of measures was calculated for each participant by dividing the total test information by the amount of items administered. To compare PROMIS measures (full item bank, short form, and CAT), the relative efficiency between measures was calculated by dividing the mean efficiency of one measure by the other. The mean (SD) T-score of the Dutch population was calculated based on the U.S. parameters. Using percentiles good (≥ 26th percentile), fair (6–25th percentiles), and poor (≤ 5th percentile) functioning cut-offs were determined, in accordance with recently defined U.S. cut-offs for this item bank (personal communication C. Forrest, data submitted).