VALIDATION DATA

The reliability of four PROMIS sleep/pain measures among participants with ME/CFS study found that all four measures showed high internal consistency with ω ranging from 0.92 to 0.97.

The average standard errors of T-scores were well below the hypothesized threshold of 3.2, ranging from 2.1 to 2.7. The study suggested that the ω coefficient and the standard error of T-score were more accurate estimates of reliability compared to Cronbach’s α. Although the α coefficients for the two sleep measures were slightly below the hypothesized threshold of 0.9, the study suggested that this may be due to a violation of the tau-equivalence assumption. Overall, the study concluded that the four PROMIS measures provided highly reliable scores for both group-level analysis and for assessing individual ME/CFS patients.

The scales were developed using IRT analyses, which allowed for the creation of both computer adaptive testing (CAT) and 8-item static short forms. The short forms were developed using CAT simulations and clinical judgment from content experts and were found to have high total test information and low standard error across a broad range of θ values, indicating strong reliability and validity. The sample used for calibration included individuals with and without sleep disorders to reflect a wide range of symptom severity. Overall, the use of IRT analyses allowed for the development of reliable and valid measures that can be customized for specific applications.

A study found that the pain short forms demonstrated an excellent fit to a unidimensional confirmatory factor analysis (CFA) model, with both the Comparative Fit Index (CFI) and Tucker-Lewis Index (TLI) exceeding 0.95. However, the fit of the sleep short forms to a unidimensional CFA model was only marginal, with CFI and TLI values above 0.90. To improve the model fit, the study conducted exploratory factor analysis (EFA) and bi-factor analyses on the two sleep short forms. These analyses showed that all model fit indices improved to acceptable ranges.

Although the Root Mean Square Error of Approximation (RMSEA) values for the pain measures were higher than the published criterion of 0.06, the authors considered them to be sufficiently unidimensional based on the high CFI and TLI values. The study also found that the RMSEA estimates from the current study were consistent with those previously reported for PROMIS measures. Overall, the study suggests that the EFA and bi-factor analyses can improve the model fit of sleep short forms, and that the CFI and TLI values can provide a useful indication of the unidimensionality of pain short forms.