For meta - analyses
of dichotomous outcomes that we include in the «Summary of findings» tables, we will express the results as absolute risks and will use high and low observed risks among the control groups as reference points.
Not exact matches
Relatively little is known about social gradients in developmental
outcomes, with much
of the research employing
dichotomous socioeconomic indicators such as family poverty.2 5 16 Thus, it is unclear whether poor developmental
outcomes exhibit threshold effects (evident only when a certain level
of disadvantage is exceeded), gradient effects (linear declines with increasing disadvantage) or accelerating effects (progressively stronger declines with increasing disadvantage) as suggested by some recent studies.17 — 19 Further, most research has examined socioeconomic patterns for single childhood
outcomes1 or for multiple
outcomes within the physical3 4 or developmental17 18 20 health domains.
For
dichotomous outcomes, we used ORs with 95 % CIs as the effect size metric when presenting the effects
of the individual studies.
The estimated effects
of continuous
outcomes are presented as a standardised effect size and
dichotomous variables as OR.
A simulation study was conducted to evaluate the power
of our design to detect treatment effects for
dichotomous outcomes.
The baseline covariates serve as adjustment for potential differences between intervention and control families that resulted from nonrandom assignment at quasi-experimental sites or selective reporting
of outcome data.29 Results
of these adjusted analyses are reported as ORs for
dichotomous variables and as differences in means for continuous
outcomes.
For randomization sites for a
dichotomous outcome and control - group prevalences ranging from 0.2 to 0.8, we found greater than 90 % power to detect a treatment effect odds ratio (OR)
of 1.5.
Generalized regression models (logistic regression for
dichotomous outcomes, linear regression for continuous
outcomes) were used to estimate the overall adjusted effects
of Healthy Steps.26, 27 These models included site variables to account for the fact that families within sites tend to respond more similarly than those at different sites.
If some primary studies report an
outcome as a
dichotomous measure and others use a continuous measure
of the same construct, we will convert results for the former from an OR to a SMD, provided that we can assume the underlying continuous measure has approximately a normal or logistic distribution (otherwise we will perform two separate analyses).
The size
of effect is described by the standardized difference (Cohen's «d») between means or proportions, and the association between treatment group and
outcomes is described by the odds ratio for
dichotomous outcomes and the correlation ratio (η) for continuous
outcomes.
Analyses
of CBCL and YSR were combined into one
dichotomous outcome (CBCL or YSR T - score
of 64 or higher) that revealed a significantly decreased risk for problems in the intervention group compared with the control group [RR = 0.11 (0.01 to 0.82)-RSB-.48 At the 15 - year follow - up, Aronen and Arajärvi found significant positive effects on YASR Total [0.37 (0.03 to 0.71)-RSB- and Internalising 0.36 [0.02 to 0.70)-RSB-.
The relations between independent predictor variables (measures
of immunological and psychological function at entry to the trial, age
of onset, and duration
of illness) and dependent
dichotomous outcome variables (self rated global
outcome; presence or absence
of caseness on the general health questionnaire at follow up; reduced or normal delayed responses to hypersensitivity skin test) were examined in separate logistic regression analyses.
As the
outcome of interest — removal from the classroom through either suspension or expulsion — was a
dichotomous variable, a logistic regression was most appropriate.
A third model with the same predictors and random effects was performed with the FaceReader measure
of disgust as
dichotomous outcome variable.
No gender differences were found with respect to attachment to mother (χ 2 (1) =.003, p >.05) or father (χ 2 (1) =.26, p >.05), nor were there any effects
of child age (entered in a logistic regression with
dichotomous attachment classification as
outcome variable) for mother B =.02, p =.67 and father B = −.03, p =.49.