Interpreting score differences in the Insomnia Severity Index: using health-related outcomes to define the minimally important difference