Longitudinal Associations Between Fathers’ Heavy Drinking Patterns and Children’s Psychosocial Adjustment

Earlier papers based on this sample have contained data on partner violence (O’Farrell et al., 2003) and on adjustment differences in children of relapsed and remitted alcoholics (Burdzovic Andreas et al., 2006), but these prior articles did not contain the data presented here regarding the associations between temporal fluctuations in fathers’ heavy drinking and children’s adjustment.

The issue of sample attrition will be considered at the end of the results section.

Results are reported with Huynh–Feldt correction for sphericity violation. Results for pairwise comparisons reported with the Holm–Bonferroni adjustment for multiple tests. The effect size r (Rosenthal, 1991)—a correlation coefficient for which r = .10 is considered a small effect, r = .30 a medium effect, and r = .50 a large effect (Cohen, 1988)—is also shown to aid interpretation.

We attempted to address the question of drug-abusing fathers to the best of our abilities, by repeating the main analyses reported on the next pages but with these 9 cases censored. The obtained results paralleled the results based on the entire sample.

For example, when we added children’s gender as a covariate to these analyses, the results showed that boys had greater total adjustment problems; F _Gender(1, 108) = 5.46, p = .021. However, gender did not predict changes in children’s adjustment over time; F _Gender×Time(2.07, 22.05) = .32, p = .74 (ns). Gender also did not interact with fathers’ drinking trajectory groups to predict either total problems in children, F _{Gender×Group} (2, 108) = .13, p = .88 (ns), or its changes over time; F _{Gender×Group×Time} (4.15, 224.05) = 1.36, p = .26 (ns). At the same time, the predictive power of fathers’ drinking trajectories on children’s adjustment and its changes over time remained unaffected, even with child’s gender statistically accounted for. Thus, addition of children’s gender to the GLM model did not inform our results beyond the basic main effect findings (of boys having greater problems) nor did it alter the principal findings of interest (of how changes over time in children’s adjustment were a function of their fathers’ drinking behavior).

In children of both SL and L&I (but not H&I) fathers, boys appeared to decline in clinical-level problems over time, whereas girls appeared not to change. Specifically, in sons of SL fathers, the rate of clinical-level problems was significantly reduced from 12 cases at baseline, to 3 cases at M12, Q _SL-Boys (3) = 21.4, p < .001. Even though the rate of clinical-level problems in daughters of SL fathers was reduced in half, from 2 cases at baseline to 1 case at M12, this change was non-significant given a small number of girls who experienced such problems to begin with, Q _SL-Girls (3) = 3.0 (ns). Similarly, the rate of clinical problems in sons of L&I fathers was significantly reduced from 8 cases at baseline, to 3 cases at M12, Q _L&I-Boys (3) = 8.03, p = .045, but the reduction from 2 cases at baseline to 0 cases at M12 in daughters was not significant, Q _L&I-Girls (3) = 4.0 (ns). The appearance of differential changes over time in clinical problems in these boys and girls was most likely an artifact of a small number of girls with such problems at baseline, and not of a real gender difference affecting the rate of change. For logistic regression analyses, we also found that child gender did not interact significantly with their fathers’ drinking trajectory to predict children’s clinical-level status at each assessment. Thus, the results are reported for the entire sample, not accounting for gender of the child.