Modified COSMIC 16
Coding was conducted by the primary M-COSMIC developer (SC) and a second
researcher (LB) who received approximately 25 hours of training in use of the instrument.
Footage collected for a child was allocated to one of the two raters for primary coding. In
conducting a formal inter-rater reliability (IRR) check, the two raters also independently double-
coded around 50% of all tapes (n = 20). While both were aware of the aims of the study, neither
was aware of the children’s scores on the standardised clinic measures. IRR was calculated using
intra-class correlation coefficients (ICCs; see Table3). ResultantICCs weregood, above0.66
(p’s < .001)andwiththemajorityabove0.84.Thiswas withtheexceptionofsome codesfor
Compliance, Action, and Follow gaze/point. However, given that high reliability was
demonstrated in one of the two activity settings (Compliance FP, r = .54; Action FP, r = .68; and
Follow gaze/point ACT, r = .89), these coding categories were retained.
- Insert Table 3 -
5.2 Association with non-verbal DQ
The majority of M-COSMIC codes were significantly associated with children’s non-
verbal DQ, with the exception of two behaviour regulation functions; show off/attention seek and
eye contact. The significant correlations ranged from r = .37 (Action) to .73 (Joint attention),
with a mean association of r = .52. As such, associations among M-COSMIC codes and other
items were sought with the effects of DQ removed. While some item distributions were
negatively skewed, results obtained through use of parametric and non-parametric statistics did
not differ. Parametric analyses are therefore presented so as to permit the more straightforward
control for DQ through the use of partial correlations.
5.3 Associations between M-COSMIC and ADOS-G