A meta-analytic approach to growth curve analysis

Area of Science:

• Statistics and Data Analysis
• Psychometrics
• Health Services Research

Background:

• Longitudinal studies often face challenges with unbalanced designs and heterogeneous treatment effects.
• Traditional growth curve analysis may not adequately address individual variability in developmental rates.
• Evaluating interventions for severe mental illness requires robust analytical methods to capture nuanced outcomes.

Purpose of the Study:

• To describe and illustrate a meta-analytic approach to growth curve analysis for longitudinal data.
• To apply this method to evaluate the Arizona Pilot Project for financing treatment of the severely mentally ill.
• To model individual differences in growth curve parameters and assess treatment effects.

Main Methods:

• Individual growth curve parameters (slopes, intercepts, residuals) were estimated for each subject using linear regression.
• A meta-analytic causal modeling approach was employed, utilizing factor analysis and general linear models.
• Exogenous predictors included method of payment, treatment site, and pretreatment assessment, with statistical controls for common factors.

Main Results:

• Two common factors (general psychological and general functional) were identified across growth curve parameters.
• Significant variations in initial status (intercepts) were found across treatment conditions, sites, and pretreatment levels.
• These variations threatened the validity of the Arizona Pilot Project's randomized design, necessitating a nonequivalent groups approach.

Conclusions:

• The meta-analytic growth curve approach accommodates unbalanced designs and individual differences effectively.
• The Arizona Pilot Project's design validity was compromised by significant baseline variations, impacting treatment outcome interpretation.
• Controlling for baseline differences revealed substantial effects of initial status on individual growth trajectories.