|
| Study | Did the study report on association between T1 cognitive, language, or adaptive scores and starting age? | How was the association of T1 cognitive, language, or adaptive scores and starting age controlled for? |
|
| Studies explicitly comparing earlier- and later-starting children on outcomes after intervention |
| Vivanti, Dissanayake, & the ASEDCC team, 2016 | The two groups of younger starting vs. older starting were compared on cognitive, verbal, and adaptive scores. Older children were found to have lower MSEL Nonverbal DQ scores | Thus, for the ANOVAs, Nonverbal DQ was entered as a covariate on all relevant analyses |
|
| Studies comparing higher and lower outcome groups on T1 variables including starting age |
| Pellicano, 2012 | The two groups of children with IQ > 80—one with ASD vs. no longer having ASD upon follow-up—were compared on cognitive and receptive language scores | Statistical analysis showed no T1 differences on PPVT-III and Leiter-R scores; in this way, these variables were held constant |
| Anderson, Liang, & Lord, 2014 | The two groups of children with VIQ > 70—one with ASD vs. no longer having ASD (“very positive Outcome”)—were compared on adaptive scores at age 2 years | Statistical analysis showed no group differences at T1 in VABS scores; in this way, the variable was held constant |
| Hedvall et al., 2015 | The two groups of children who had gained the most vs. lost the most from T1 to T2 were compared on adaptive and language scores at T1 as well as age at referral. Statistical analysis showed significant differences in VABS and MCDI scores, as well as age at referral | Logistic regression controlled for covariation of predictors |
|
| Studies that added or focused on prediction during or after evaluation of an intervention |
| Itzchak & Zachor 2011 | No | Regression analysis controlled for all other T1 variables including cognitive and adaptive scores |
| Perry et al., 2011 | No | Regression analysis controlled for all other T1 variables including cognitive and adaptive scores |
| Flanagan, Perry, & Freeman, 2012 | No | Sequential regressions entered intervention duration, age, and T1 adaptive skills, as well as other predictors and interaction terms, e.g., Age × Group |
| Kasari et al., 2012 | No | First a forward regression procedure was used to identify strong predictors. Then hierarchical regression used predictors of age at the first assessment and play level, among others |
| Rogers et al., 20121 | No | Linear regression analysis tested age of starting EI program controlling for all other predictors, including T1 cognitive, ADOS Social-Affect, and Imitation |
| Smith, Klorman, & Mruzek, 2015 | No | Sequential multiple regression analysis entered T1 scores for the cognitive and adaptive outcome measure under analysis and age at intake. The interaction of Time × Age was also assessed, with effect of cognitive scores accounted for |
| Eapen, Crncec, & Walter, 2016 | No | Linear regression analyses were conducted using predictors of T1 cognitive, adaptive, and play scores |
| Vivanti et al., 2018 | No | First, a set of partial correlations were examined between age of entry and each outcome variable, partialling out the Time 1 score for each variable. Next, a linear regression entered age of starting the intervention after the variance associated with baseline Verbal DQ and treatment group was accounted for |
|
| Trajectory analysis studies |
| Virues-Ortega, Rodriguez, & Yu, 2013 | No | Multilevel regression analyses were used to predict an established growth curve, entering T1 scores to establish which accounted for the best fit and then which additional predictors added significantly to the goodness of fit. Predictors included pretreatment functioning level and age, among others |
| Tiura, Kim, Detmers, & Baldi, 2017 | Yes, since they examined predictors for Time 1 cognitive levels to find that age of entry was significantly related. Children who were older when starting had higher cognitive levels | Multilevel regression analyses were used to predict an established growth curve, entering four T1 predictors to establish, first, which predicted Time 1 functioning levels, and second, which accounted for the best fit for the growth curves |
|
