본문 바로가기
읽기 · 학술 · 공개 라이선스 원문

Parenting interventions to promote early child development in the first three years of life: A global systematic review and meta-analysis

Jeong, Franchett, Ramos de Oliveira, Rehmani, Yousafzai · 2021

공개 학술지 PLoS Medicine 에 실린 총설 논문입니다(2021년). CC BY 4.0 으로 공개된 글이고, 그림 · 표 · 참고 문헌 표시를 뺀 본문을 실었습니다.

출처 Jeong et al. (2021), PLoS Medicine · 라이선스 CC BY 4.0 · 원문

장 4개 올림 이 장 1184낱말
← 앞 장 다음 장 →

이 장은 원문만 있습니다. 문장별 해석은 앞 장부터 차례로 붙입니다.

2장 · Parenting interventions to promote early child development in the first three years of life: A global systematic review and meta-analysis — Methods

Search strategy and selection criteria

This systematic review and meta-analysis was conducted and reported using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (S1 PRISMA Checklist). Six electronic bibliographic databases (MEDLINE, Embase, PsycINFO, CINAHL, Web of Science, and Global Health Library) were searched for articles published from database inception until May 15, 2019. The search was subsequently updated during the peer-review process to include articles published between May 16, 2019 to November 15, 2020. The search strategy (S1 Text) was informed by search terms and keywords used in prior systematic reviews related to parenting interventions and through consultations with a research librarian. Reference lists of relevant studies and reviews were also scanned for any additional studies that may have been missed.

Full-text, peer-reviewed articles were included if they met the following criteria: (1) parenting interventions that aimed to improve interactions, behaviors, knowledge, beliefs, attitudes or practices of parents with their children in order to improve ECD; (2) evaluated using a randomized controlled study design; (3) targeted children and their parents during early childhood (pregnancy through the first 3 years of life); and (4) measured at least 1 ECD outcome after the completion of the intervention (or shortly thereafter). In this study, parents are broadly defined as the legal guardian, biological parent, or adult caregiver responsible for the well-being of the child. Studies were excluded if they met any of the following criteria: (1) not a relevant parenting intervention focused on promoting ECD; (2) nonrandomized study design (e.g., quasi-experimental studies); (3) targeted a population of children who were, on average, older than 36 months; (4) targeted a population of children or parents who had a diagnosed illness or disability; and/or (5) did not measure at least 1 ECD outcome.

Titles and abstracts of all identified articles were independently screened by at least 2 reviewers (JJ, EF, and a team of graduate research assistants). If the title and abstract screening was insufficient for determining eligibility, the full-text article was reviewed for eligibility criteria. Any discrepancies between the 2 reviewers regarding the eligibility of a study were resolved through discussions among the reviewers, with input from another author (AKY) as needed until consensus was reached.

For studies meeting the eligibility criteria, full-text articles were reviewed. Data were extracted by JJ, EF, and a team of trained graduate research assistants using a prepiloted extraction form. JJ and EF trained the graduate research assistants over the course of a 3-week training period on how to use the data extraction sheet based on pilot exercises of relevant articles. Throughout the entire data extraction process, each data extraction completed by the research assistants was independently extracted by a second reviewer (JJ or EF), and any issues were discussed and resolved during the weekly in-person team meetings with the research assistants. For additional quality assurance, 35% of assigned articles were randomly selected for independent data extraction by a third reviewer (JJ or EF). Any discrepancies between the master reviewers (JJ and EF) regarding the extracted data were resolved through discussions and consensus together with an additional reviewer (AKY).

Outcomes

The primary outcomes were 6 domains of ECD: cognitive, language, motor, and socioemotional development, behavior problems, and attachment. These ECD outcome domains have been prioritized in previous work, commonly measured across global contexts, and are each unique aspects of child development during the first 3 years of life. Moreover, each of these ECD outcome domains has shown predictive validity with later achievement, function, performance, or outcomes. Measures of developmental disabilities (e.g., autism or ADHD) did not qualify as ECD outcomes. The secondary outcomes were 4 measures of parenting: parenting knowledge, parenting practices, parent–child interactions, and parental depressive symptoms.

Data analysis

Effect sizes on ECD and parenting outcomes were calculated as SMD between the intervention and comparison arms (e.g., no intervention, nonparenting intervention, standard of care), with respect to change in mean values from baseline to endline (when baseline values were reported) after standardization by their pooled standard deviation (SD). In multiarm studies, the comparison group was typically the intervention arm without any parenting components and/or the standard of care. For studies that reported outcomes from multiple follow-up waves of assessments, outcome measurements from the time point closest to the completion of the intervention were used for the analyses. The magnitude of effect sizes was interpreted in the context of public health, pediatric, and early education interventions and the practical significance with respect to the outcomes.

Each of the primary and secondary outcomes was examined in separate models. We used a robust variance estimation meta-analysis model to account for multiple measurements per trial per outcome domain. This approach allows for the inclusion of any number of dependent effect size estimates from a given study to contribute to the pooled effect size estimate for an outcome, resulting in increased power and more precise estimates. For example, this approach was used to handle the separately reported effect sizes for receptive language and expressive language in a given study for overall language development or both internalizing and externalizing behaviors for overall child behavior problems. Pooled effect size estimates were based on random effects models. P values < 0.05 denoted statistical significance. Heterogeneity of the pooled effect size was assessed using the I² statistic and the Cochran Q test and its P value.

Moderator analyses were conducted on all outcomes using random effects meta-regression models to explore potential sources of heterogeneity in the average effect by the following prespecified sample, intervention, and study characteristics: country-income level (LMICs or high-income countries (HICs)), average child age at baseline in months, whether the parenting intervention included a component to enhance responsive caregiving (see S1 Table for classification of responsive caregiving interventions), intervention duration in months, delivery modality (individual, group, or both), setting (home, community, clinic, or combination), and study quality score. Study quality was assessed using the Cochrane Collaboration Risk of Bias Assessment Tool for randomized controlled trials (RCTs). Studies were rated as low risk (0 points), unclear risk (1 point), and high risk (2 points) for a total risk of bias score ranging from 0 to 16. For the continuous moderator variables, we generated a binary variable for child age at baseline and intervention duration (< or ≥12 months) and a binary study quality variable based on a mean split of the total risk of bias score (lower or higher risk of bias) and compared effect estimates according to these subgroups. Statistical significance for the moderator variables of country-income level, intervention content, delivery, and setting was determined from the Qbetween test statistic which was evaluated on the χ² distribution. For the remaining moderator variables that were originally continuous measures (i.e., child age, intervention duration, and study quality score), statistical significance was determined from meta-regressions using the continuous measure (e.g., child age in months).

Publication bias was examined for each outcome analysis using the Egger’s regression test to test the null hypothesis of small-study bias (i.e., the extent to which less precise or smaller sample studies yield greater effects than larger sample studies). All analyses were conducted using Stata version 16. This review was preregistered with PROSPERO (CRD42018092458 and CRD42018092461). See S1 Protocol for the final review protocol used for the present study and notes regarding any changes made from the initial prospectively registered protocol.

이 글은 위 출처의 원문을 그 라이선스(CC BY 4.0)에 따라 옮겼습니다. 참고문헌 목록과 인용 번호, 그림 · 표 · 각주 · 수식 · 부록은 빼고 절별로 나누어 실었으며, 그 밖의 문장은 원문 그대로입니다. 원 저작자와 발행 기관은 이 화면을 만들거나 검토하지 않았고, 이곳의 해석과 낱말 풀이는 공식 번역이 아닙니다.