Page 65 - Post-traumatic stress disorder in adults with mild intellectual disability: Screening, assessment, and brief intensive EMDR therapy

Development and evaluation of TS-ID
Analyses
To determine the reliability of the adult and proxy versions of the TS-ID, the
internal consistency of the total symptom frequency score (section two)
was calculated using Cronbach’s alpha (SPSS version 27). The validity of
the TS-ID adult version and TS-ID proxy version was assessed by comparing
the total symptom frequency score of both versions with the final outcome
of the DITS-ID (i.e. presence or absence of PTSD) using receiver operating
characteristic (ROC) analysis. The discriminative capacity of both versions
of the TS-ID was operationalised by calculating the AUC, which reflects their
ability to distinguish between individuals with and without PTSD. When an
optimal AUC value was obtained, we determined a cut-off point based
on the optimal ratio between sensitivity (justified positive prediction) and
1 − specificity (false-positive prediction). In addition, Youden’s J (Youden,
1950) was used as a supplementary evaluation to assess overall
discriminative power. Furthermore, the positive predictive value (PPV) and
negative predictive value (NPV) were calculated to provide additional insight
into the predictive accuracy of the different cut-off points. In determining
the cut-off score, explicit consideration was given to the outcomes of the
expert meeting of professionals who will use the TS-ID in clinical practice.
The number of participants required was calculated using the R package
for power calculation in diagnostic tests (Chernick & Liu, 2002; Chu & Cole,
2007; Flahault et al., 2005). Assuming a sensitivity and specificity of .8, an
estimated precision of .2 (delta .2), a significance level of .05 and a power of
.8 in a sample in which the distribution of yes/no PTSD is equally distributed
(prevalence .5), a total of 78 participants were needed. If more than 10% of
the questions in the DITS-ID, TS-ID adult version or TS-ID proxy version were
not scored, the questionnaire was not included in the analyses. Accordingly,
one TS-ID proxy questionnaire was excluded.
3
63