The first methodological hypothesis was that studies published more recently would show fewer negative and more positive outcomes than earlier studies. Lamorey (1999) observed more recent studies to show fewer negative effects and more variation in outcomes. A second methodological hypothesis was that more negative effects would be found for parental reports than sibling self reports (Summers et al. , 1994).
The third methodological hypothesis was that studies employing normative data for comparison to the sibling samples would produce negative effects of greater magnitude than found for studies that employed matched control groups (Lavigne & Faier-Routman, 1992). Substantive Issues. A number of hypotheses were also made that related to substantive variables. First, a larger negative effect was expected for internalizing over externalizing behaviors.
Howe (1993) found four of eleven studies of siblings of children with chronic illness showed a negative effect for internalizing behavior compared to only one of eight studies for externalizing behaviors. Second, sibling outcomes were anticipated to vary by the chronic disease and its features. One view is many chronic conditions of childhood produce similar psychological and behavioral effects (Vessey & Mebane, 2000).
Childhood chronic illnesses, however, vary on dimensions such as etiology, age of onset, impact on functioning, and prognosis (see Lobato, Faust, & Spirito, 1988). More severe chronic illnesses place greater restrictions on the child’s activities (Newacheck & Taylor, 1992), and perhaps greater demands on parents, siblings, the family system, and the community (Patterson, 1988). Third, the interaction of sibling gender and birth order was considered (Howe, 1993; Williams, 1997). Method
Fifty published studies from 1976 to 2000, representing over twenty-five hundred siblings of children with chronic illness, were identified from computer searches of databases such as PsycLit and MEDLINE, using key words such as “siblings” and “illness,” from previous reviews of the literature and from the reference sections of located studies. Excluded from the meta-analysis were case studies, nonempirical or qualitative studies, or studies without an appropriate comparison group or normative data.
Studies were also excluded that evaluated the reactions of healthy siblings to the illness or death of a brother or sister or pertained to the adult siblings of individuals with a chronic illness. Studies that employed no comparison group but that provided normative data were included in this meta-analysis. Unpublished studies were not sought for inclusion in this meta-analysis. First, it is almost impossible to collect all published studies in all languages, much less all unpublished studies.
Second, the peer review process for published studies serves as an albeit imperfect form of quality control. Third, there is evidence that publication bias is less serious than once feared (Sharpe, 1997). Publication bias, the socalled “file-drawer” problem, is the belief that the failure to include unpublished studies in the metaanalysis might inflate the magnitude of effect sizes, given that published studies may overrepresent statistically significant findings.
To ascertain the likelihood of such a publication bias, statistical and graphical analyses of effect sizes were conducted. Studies by the same author(s) that appeared to examine the same participants (e. g. , Breslau & Prabucki, 1987; Breslau, Weitzman, & Messenger, 1981) were treated as a single study for the purposes of this meta-analysis. Three of the primary studies (Faux, 1991; Stawski, Averbach, Barasch, Lerner, & Zimin, 1997; Wood et al. , 1988) provided separate data for the siblings of children with distinctly different chronic illnesses.
These sub samples were treated as separate studies. In total, 51 study-level effect size statistics were evaluated. Each study was coded for method of data collection (child self-report, parent report, or direct observation), chronic illness, age of siblings, gender of siblings, number of sibling and comparison participants, and dependent measure category: psychological functioning (e. g. , Internalizing subscales of the Child Behavior Checklist), self concept (e. g. , Piers-Harris Self-Concept scale), caretaking, sibling relationship, peer activities (e.
g. , Social Competence subscale of the Child Behavior Checklist), cognitive functioning (e. g. , intelligence test scores), and cognitive development (e. g. , school performance). Parent and teacher reports were combined because only five studies asked teachers to complete a dependent measure. Four of the five comparisons based on teacher reports were not statistically significant. All codings were completed by the first author and checked independently by the second author. Disagreements were resolved by discussion.