Some non-randomized constructed comparison groups for evaluating age-related outcomes of intervention programs

The evaluations of many intervention programs for young children are hamper ed by the lack of untreated comparison groups. Simple pre-post comparisons or more elaborate growth profile analyses within a single treatment group c annot separate changes associated with the treatment from normal developmen tal changes typical of the participant population. To address this issue, s ome non-randomized constructed comparison strategies that can be used to ev aluate interventions are outlined and illustrated for outcome variables tha t would be expected to change over age if no treatment were given. Such a d ependent variable might include general indices of developmental progress i n young children, which may decline over the first few years of life in low -income, high-risk samples or may increase over age in middle-class groups suffering adverse perinatal circumstances and disorders even without specia l interventions. The proposed strategy consists of determining an expected age function for the dependent variable using pretest scores (if participan ts enter the treatment program at different ages) or using pretest scores o f older siblings who vary in age at the start of the program, and then calc ulating an age-adjusted expected posttest score against which actual postte st performance may be evaluated. This approach also permits analysis of par ticipants enrolled in the treatment for varying lengths of time and over so mewhat different age spans.