Q-TOP/ETRAD-Q

This caregiver-report questionnaire comprises 42 items regarding children’s behaviours to be rated on a four-point Likert scale (not at all true, a little bit true, pretty much true, very much true). The questionnaire comprise a RAD scale (25 items) and a DSED scale (17 items*).

There are two subscales for DSED:

• Interaction with unfamiliar adults (criteria A1, A4): 8 items
• Social disinhibition (criteria A2, A3 and B): 9 items

There are three subscales for RAD:

• Low selective attachment (criteria A1 and A2): 8 items
• Low social and emotional responsiveness (criteria B1 and B2): 9 items
• Emotional unpredictability (criterion B3): 8 items

The main validation study (Monette et al., 2022) published in Assessment showed that the ETRAD-Q possesses excellent psychometric properties in terms of factor validity, X²(774) = 2102.827, RMSEA = .046, CFI = .959, TLI = .957, SRMR = .061, when tested on a sample of community children (n = 578) and at-risk children (n = 245 in out-of-home care or adopted), and in terms of internal consistency (all Omega > .88) and test-retest reliability (all r > .83). The ETRAD-Q also demonstrated very good convergent validity. Its RAD scale correlated strongly with the RAD scale of the RPQ, (r = .83), and of the RADA interview (r = .75); and its DSED scale correlated strongly with the DSED scale of the RPQ (r = .89), and of the RADA interview (r = .78).

The ETRAD-Q subscales allow differentiating three groups of children:1) Community children, 2) At-risk children (adopted or in CPS care), 3) At-risk children with a clinical diagnosis of attachment disorder (by a community practitioner).

The technical manual (Monette, 2024) confirms that the ETRAD-Q exhibits excellent psychometric properties, confirmed by analyses conducted on the Wave 2 sample (n = 789), which is composed of community children and at-risk children.

The scale reliability is very high, as measured by McDonald’s Omega. For the full version, the TRA scale (Reactive Attachment Disorder) reached .94 and the DSED scale (Disinhibited Social Engagement Disorder) reached .93. The sub-scales also maintain strong internal consistency, with values ranging from .86 to .89. The short form (16 items) also demonstrates strong internal consistency (RAD-SF: .79; DSED-SF: .89). The Confirmatory Factor Analysis (CFA) replicated on Wave 2 supports the hypothesized structure: a model with two second-order factors and five first-order factors for the full version, showing adequate fit indices (RMSEA = .064, CFI = .919). For the short version, the model with two distinct factors (RAD and DSED) is confirmed (RMSEA = .082, CFI = .955), and all factor loadings are strong (> .50).

Convergent validity is established by high correlations between the ETRAD-Q scales and those of reference instruments. Correlations between the DSED scale of the ETRAD-Q and the DSED factors of the RPQ and the RADA are .90 and .79, respectively. For the RAD scale, correlations are .77 (with the RPQ) and .71 (with the RADA). The ETRAD-Q scales also show moderate to high correlations (from r = .37 to r = .75) with the level of functional impairment in the child, as measured by the BFIS-CA.

Area Under the Curve (AUC) analyses, conducted on a combined sample (Waves 1 and 2) using classification by the RADA Interview as the criterion, indicate excellent discriminatory capacity for the ETRAD-Q. The full-version DSED scale reached an AUC of .91, which is considered exceptional, while the RAD scale reached .88, judged excellent.

In terms of known-group validity, community children generally obtain lower scores on the ETRAD-Q scales than children followed by Child Protection Services (CPS) who do not have a TRA or DSED diagnosis (via the RADA interview), and the latter group obtains lower scores than children followed by CPS who have received a RAD or DSED diagnosis. Children with ASD (Autism Spectrum Disorder) typically show scores comparable to those of SPJ children (without a TRA or DSED diagnosis).

The ETRAD-Q is a screening tool, and cut-off scores have been suggested using the Youden criterion to maximize both sensitivity and specificity. The cut-off for the full-version DSED scale is set at 15/51 (87th percentile) and for the RAD scale at 34/75 (91st percentile). The use of these thresholds ensures a very high Negative Predictive Value (NPV) of 99% for both scales, indicating that the absence of symptoms according to the ETRAD-Q is highly reliable. However, the Positive Predictive Value (PPV) is low (34% for DSED, 13% for RAD), highlighting a high rate of false positives and reaffirming that the ETRAD-Q must be followed by an in-depth diagnostic evaluation. Finally, the clinical norms (T-scores) are established on a final normative sample (n = 662) which was trimmed to better represent the educational attainment of the Quebec population. Given the significant positive skewness and non-normality of scores in the general population, the T-scores are derived from the observed percentile ranks rather than the mean and standard deviation.

The technical manual (to be downloaded with the ETRAD-Q kit, see below) includes  normative data. Reading the technical manual is required to use the tool. We would like to thank the Institut universitaire Jeunes en difficulté (IUJD) for the financial support that made it possible to complete the validation study, as well as the development of the ETRAD-Q kit (web page, technical manual, final questionnaire, Excel corrector). The kit is available in French and English. If you would like to translate the instrument into another language, please contact us. There are already teams working on translating the questionnaire into different languages at the moment.

* The validation paper (Monette et al., 2022) suggested removing one item from the DSED scale, however, additional validation analyses in the technical manual (Monette, 2024) support the use of all questionnaire items (42 items).