Goal attainment scaling (GAS) is a patient-centric outcome measure that captures meaningful change through personally identified treatment goals. For each goal, a 5-point attainment scale is defined to describe possible outcomes, with the midpoint representing the “expected outcome”. A key assumption in GAS analysis is that the scores on the scale approximate a normal distribution. Using data simulation techniques, we investigated whether GAS performance varied if the assumption of normality was violated.

The statistical power did not vary significantly whether using uniform or normal intervals, with less than 1% difference for most parameter combinations. As expected, T-scores were higher when using uniform intervals. However, the standardized effect sizes were equal across small-medium effect sizes (<0.7), as typically seen in most GAS studies. At the higher effect sizes, the uniform intervals underestimated the treatment effect compared to normal intervals (difference of 0.03-0.07). This difference is unlikely to impact interpretation of GAS results.

Violating the distributional assumptions of GAS data is unlikely to affect the power to detect a treatment effect at effect sizes typically seen in GAS studies. The GAS method is robust to perturbations in normality of goal scores.