2.4. Coding of Primary Studies

To extract relevant information from the selected primary studies, an Excel spreadsheet was prepared. All the included studies were coded independently and simultaneously by two independent raters (the percentage of agreement was 95%). Discrepancies were discussed with a third rater and solved among the three evaluators.

First, the characteristics of the publication were coded: type of publication (i.e., journal article or grey literature), year of publication, and the language of publication. Second, the characteristics of the studies were coded: funding sources (i.e., international funding, national funding, local funding, multiple funding sources); the number of waves of the longitudinal design; the interval between waves; the dimensions of each study; and the source of information used to evaluate them (i.e., self-reports, objective assessment). Third, the characteristics of the participants were coded: sample size, gender composition of the sample (% females), mean age, geographical location, and ethnic composition of the sample.

Finally, data necessary for effect size computations were extracted. Due to the high heterogeneity of the studies included, different effect sizes were coded (i.e., odds ratio, cross-lagged correlations, Spearman’s rho, beta coefficients) to address how family factors (demographic and relational) and sleep quality indicators were longitudinally related (see Section 2.5). If only standardized beta regression coefficients were reported, the correlation coefficients were estimated based on Peterson and Brown’s formula [44]. When data for effect size computations were not reported in primary studies, study authors were contacted by email to request missing data. In total, nine authors were contacted to obtain all (or part of) the necessary data for effect size computations. If authors did not answer the first request, three reminders (one every two weeks) were scheduled. Two authors replied by providing the requested data; five replied specifying that they could not provide the required data (e.g., they could not access the dataset anymore); and two did not respond to the request. For this reason, seven studies were excluded because there were insufficient data as indicated in the PRISMA diagram (full texts excluded because of missing data; Figure 1).