Abstract

Linked Article
BACKGROUND
Description of the condition
Freedom from abuse, neglect and violence is a basic right of all children (
In most high-income countries, primary responsibility for child protection rests with legislated child protection agencies, either operated or overseen by the state. While the specific mandate of these agencies differs across jurisdictions, statutory responsibilities usually include: receiving and investigating reports of suspected abuse and neglect; providing ongoing protection services to families in which children have been deemed at risk of future maltreatment; and the provision of out-of-home care when children cannot be safely cared for at home. While some families or children self-refer to a child protection agency, the majority of families referred are “involuntary” insofar as they have not asked for nor consented to the referral.
Traditionally, decision-making by child protection agencies has been professionally-driven, with workers conducting assessments of families’ problems and risk profiles, and determining a treatment plan with which families are asked to comply (
Within the last two decades there has been a growing acknowledgement of the need to work more collaboratively with families in the provision of child protection services (
Description of the intervention
Family Group Decision Making is an umbrella term [1] for practice models that shift planning for children away from “professionally driven” towards a more “family-centered” approach, with the premise that families are experts on their own situations, and as such, should be considered well qualified to contribute to plans designed to promote the safety and well-being of their children (
There are numerous models of FGDM used internationally, including Family Group Conferencing (FGC), the Family Unity Meeting (FUM) model, Team Decision Making, Family Team Meetings, and Family Team Conferencing (
Family Group Conferencing (FGC) was created in New Zealand through a collaborative effort between Maori[1], governmental and community leaders, based on traditional Maori decision-making processes (
For the purposes of the current review, family group decision making interventions include all models that involve the convening of family and extended family members, identified friends and/or community members; child protection professionals; and, if needed or requested, other professionals, in an effort to collaboratively develop a plan to maintain child safety, facilitate stable and permanent living arrangements, and promote child and family well-being. The Maori are the indigenous people of Aotearoa New Zealand, They comprise approximately 14% of New Zealand's population, living mostly on New Zealand's North Island ( The Maori concept of family differs from the traditional Western construct. The Maori words whanau, hapu and iwi, while not readily translatable, have a range of meanings from extended family to tribal affiliation, and comprise the familial structure upon which Maori society is based (
How the intervention might work
While the theory of how FGDM processes lead to specific client outcomes remains both understudied and underdeveloped (
Plans resulting from FGDM are also believed to be more consistent with the child and family's cultural beliefs and identities (
Finally, FGDM models frame families as competent and often explicitly focus on their strengths, with the aim of empowering families and shifting their experience of child protection service from one characterized by powerlessness to one of self-determination and collaboration (
Why it is important to do this review
To date, FGDM models have been widely implemented in several countries, including New Zealand, the U.K., Canada, the United States, Australia, France, South Africa, Sweden, Norway, Denmark, Israel and the Netherlands (
Family Group Decision Making models are conceptually compelling and consistent with social work values and principles of empowerment and culturally appropriate practice. These models have been widely implemented internationally in child welfare contexts. In addition to the precedent-setting New Zealand legislation, several countries and jurisdictions have legislation or policies encouraging the use of FGC or FGDM in cases of child abuse and neglect, and provide government funding for FGDM programs.
Despite the widespread support and investment in FGDM, key outcomes for children and families who receive FGDM interventions (safety, permanence and well-being) are not well documented, particularly over the longer term (
OBJECTIVES
To assess the effectiveness of the formal use of FGDM in terms of child safety, permanence (of childs living situation), child and family well-being, and client satisfaction with the decision-making process.
METHODS
Criteria for considering studies for this review
Types of studies
Studies will be eligible for this review if they: 1) used random assignment to create treatment and comparison or control groups; or 2) used parallel cohort designs in which groups were assessed at the same points in time (i.e., quasi-experimental designs that include groups assessed at the same time as opposed to a historical cohort). Single-group designs and single-subject designs will be excluded (see ‘risk of bias’ section for further details on included designs).
Types of participants
Children and young people aged 0-18 years who have been the subject of a child maltreatment investigation.
Types of interventions
Any form of Family Group Decision-Making (FGDM) used in the course of a child maltreatment investigation or during the course of services arising from such an investigation.
This involves convening family, extended family, identified friends and/or community members along with child protection professionals and other professional, community-based collaterals in an effort to collaboratively develop a plan to maintain child safety, facilitate stable and permanent living arrangements, and promote child well-being. Therefore, studies will be included in the review if they involve: 1) a concerted effort to convene family, including extended family, friends and community members; and 2) child protection professionals (as well as other professional service providers) participating in; 3) a planned meeting with the intention of working collaboratively to develop a plan for the safety, permanence and well-being of child(ren); and 4) with a focus on family-centred decision-making.
Types of outcome measures
Primary outcomes
Substantiated/verified/indicated referrals to a child protection authority (gold standard) Referrals (not verified) to a child protection authority Parent self-report Child self-report Collateral party report
Analyses may include all of these types of measurements, but these will first be grouped by indicator, ‘best’ source will be preferred, and studies will be analyzed separately prior to synthesis.
For children residing in the homes of their birth parents, entry into foster care or other out-of-home placement will be interpreted in a two different ways. First, any and all placements will be analyzed as a negative outcome (i.e., FGDM, in this instance, is being used to prevent any placement, including kinship placements, into out-of-home care). However, kinship care (placement in out-of-home care with relatives) is a unique type of care that FGDM may actually faciliate as an alternative to placement to other forms of out of home care. Therefore, Kinship care will also be interpreted as a positive outcome and analyzed separately (i.e., does the rate of placement into kinship care differ between children receiving FGDM versus children who do not), and the interpretation of this set of analyses will be made within the context of FGDM's stated purpose (i.e., involving family members in child placement decisions). For children residing in out-of-home care (i.e., foster care, kinship care, group care), legal permanence will be interpreted as a positive outcome (i.e., reunification with birth parents, adoption by related or non-related caregivers, placement with relative caregivers, legal guardianship / legal custody by related or non-related caregivers (i.e., FGDM, in this instance, is being used to facilitate family permanence). For children residing in out-of-home care, long-term foster or kinship care arrangements with the same caregiver (i.e., placement stability) will also be considered positive outcomes, though such outcomes are generally held in low regard when compared to legally permanent homes.
Since the primary outcomes listed here are generally events rather than perceptions or subjective impressions, the gold standard or ‘best’ indicator for the measurement of primary outcomes will be official reports found in administrative data and case files. For prevention of child maltreatment, there may be some studies that utilize self-report or a report from a secondary party. In such cases, only studies using standardized tools measuring the occurrence of child maltreatment and/or family violence, for example the Conflict Tactics Scale (
Studies will only be included in the analysis of primary outcomes if subjects are followed for at least six months after the intervention to allow for sufficient time to observe outcomes. For included studies, longest common follow-up will be used but separate analysis of short-term (e.g., 6 months) and long-term (e.g., 3+ years) outcomes will be conducted for each primary outcome measure.
Secondary outcomes
Secondary outcomes include child
Search methods for identification of studies
Both published and unpublished work will be considered eligible for the review. A Trial Search Coordinator (Carmen Logie) will be responsible for coordinating this activity. To the greatest extent possible, the search will not be restricted to any single language or nationality.
A Systematic Information Retrieval Coding Sheet (SIRC) has been developed (see The date(s) of the search; The name of the researcher; The database used for the search; The specific search terms used in combination (including limiters and expanders); and The number of results for each search strategy.
Such recording facilitates replication of the search strategy. Furthermore, the search strategy will be saved and “copied and pasted” into the review to avoid editing errors.
Electronic searches
Electronic searches for the identification of appropriate studies will include the following bibliographic Databases: Cochrane Central Register of Controlled Trials (CENTRAL) MEDLINE The Campbell Collaboration Register of Controlled Trials (C2-SPECTR) PsycINFO EMBASE Database of Abstracts of Reviews of Effects (DARE) ASSIA (applied social sciences) ERIC CINAHL International Bibliography of the Social Sciences Caredata (social work) Social Work Abstracts Social Sciences Abstracts Child Abuse and Neglect (CANDIS) Australian Family and Society Abstracts Database Dissertation Abstracts International (DAI)
To ensure maximum sensitivity and specificity, subject headings and word text will be searched in a systematic process.
The search that will be used for Medline is as follows and will be modified accordingly for the other databases listed: family group.tw. family decision.tw. family decisionmaking.tw. family conferenc$.tw. family unity.tw. family team.tw. group conferenc$.tw. group decision.tw. group decisionmaking.tw. team conferenc$.tw.. team decision.tw. team decisionmaking.tw. or/1-12 exp child/ adolescent/ exp infant/ (child$ or adolescen$ or boy$ or girl$ or infant$ or toddler$ or baby or babies or preschool$ or pre-school or teen$).tw. or/14-17 13 and 18
Searching other resources
Reference lists
Reviewers will check the reference lists of all relevant articles that are obtained, including those from previously published reviews. Potentially relevant articles that are identified will be retrieved and assessed for possible inclusion in the review.
Personal communication
Face-to-face discussions at meetings, emails, requests on list-servs, and formal letters of request for information from authors, presenters and experts will be solicited to assist the review team to locate relevant studies. A list of the inclusion criteria for the review, along with a sample of relevant articles, will be sent to these key informants along with the request for studies. The list of experts to be contacted will include principle investigators of eligible studies, program developers, and authors of previous reviews of relevant literature.
Handsearching journals
Journals relevant to child maltreatment will be handsearched by trained researchers to uncover relevant studies not found by electronic database searches. In addition, trained reviewers will search reference lists of relevant articles. These include: Child Welfare Children and Youth Services Review Social Service Review Child Maltreatment Child Abuse and Neglect Journal of Social Services Research Social Work Research on Social Work Practice Social Work Research Child Abuse Review
Grey Literature
Special attention will be made to search and collect relevant studies captured in the grey literature. Specifically, the review will include the following strategies to locate articles:
1) Conference Proceedings (e.g. PapersFirst and ProceedingsFirst, both accessed through the University of Toronto library system); 2) Research Reports (e.g.
Data collection and analysis
Selection of studies
Titles and abstracts of studies yielded by the searches will be independently screened by two reviewers to determine their eligibility for inclusion in the review. The screening of the studies will be carried out by a three-stage procedure (see
Data extraction and management
Two review authors, using a data extraction form, will independently extract data on participants, methods, interventions, outcomes, and results (
Details to be extracted will include: Study: information regarding the author(s); year of publication; source; country; and language Characteristics of Setting and Participants: eligibility criteria for participants; explanation of recruitment procedures, setting (country, location, clinical/non clinical); demographic features of the sample Sampling: sample sizes for treatment and control; whether power analysis was used to determine sample size; allocation to the treatment and control; explanation of method used to generate the allocation Research Design: Type of design including major features such as random selection, random assignment, and non-equivalent control group. Features will be assessed according to ‘Assessment of risk of bias’ categories as described below Intervention Data: nature of interventions (for treatment and comparison/control groups); FGC, FUM, or some other form of Family Group Decision-Making; aim of intervention; length of intervention, whether manuals were used, whether fidelity checks were included, information on possible contamination reported Outcome Data: primary and secondary outcomes, measures used, information on reliability/validity of measures Results: attrition at post intervention and follow-up; number excluded from the analysis; length of follow-up; statistical methods; type of data effect size is based on; data needed for effect size calculations.
Assessment of risk of bias in included studies
Risk of bias will be assessed independently by two review authors according to the Cochrane Collaboration Handbook (
Reporting bias within each of the included studies (e.g. selection, measurement, attrition bias) will be reported in the results and discussion.
Sequence generation
The method used to generate the allocation sequence is described in detail so as to assess whether it should have produced comparable groups
Review authors’ judgment: was the allocation concealment sequence adequately generated?
Ratings: ‘Yes’ (low risk of bias); ‘No’ (high risk of bias) and ‘Unclear’ (uncertain risk of bias)
Allocation concealment
The method used to conceal allocation sequence is described in sufficient detail to assess whether intervention schedules could have been foreseen in advance of, or during, recruitment; review authors’ judgment: was allocation adequately concealed?
Ratings: ‘Yes’ (low risk of bias); ‘No’ (high risk of bias) and ‘Unclear’ (uncertain risk of bias)
Blinding
Any measures used to blind participants, personnel and outcome assessors are described so as to assess knowledge of any group as to which intervention a given participant might have received.
Review authors’ judgment: was knowledge of the allocated intervention adequately prevented during the study?
Ratings: ‘Yes’ (low risk of bias); ‘No’ (high risk of bias) and ‘Unclear’ (uncertain risk of bias)
Incomplete outcome data
If studies do not report intention-to-treat analyses, attempts are made to obtain missing data by contacting the study authors. Data on attrition and exclusions are extracted and reported, as well as the numbers involved at measurement period (compared with total randomized at pre-test), whether the reasons for attrition/exclusion are reported or obtained from study authors, and whether the study authors perform any re-inclusions of missing data in analyses.
Review authors’ judgment: were incomplete data dealt with adequately by the reviewers?
Ratings: ‘Yes’ (low risk of bias); ‘No’ (high risk of bias) and ‘Unclear’ (uncertain risk of bias)
Other sources of bias
Was the study apparently free of other problems that could put it at a high risk of bias (These will be determined once the included studies are considered, but will likely include contamination as this is present in at least one known study)
Ratings: ‘Yes’ (low risk of bias); ‘No’ (high risk of bias) and ‘Unclear’ (uncertain risk of bias)
Measures of treatment effect
For binary outcome data, effect sizes will be calculated as odds ratios (OR) with 95% confidence intervals. Continuous data will be converted into standardized mean differences (SMDs) and presented with 95% confidence intervals. When necessary, we will use formulas suggested by Lipsey and Wilson (2001) to convert correlation coefficients, F ratios, t-values, and chi-square values into SMDs. Hedges’ g will be used to correct for small sample bias.
We assume that there will be unexplained sources of heterogeneity across studies; hence assumptions of the fixed effect model (that all studies provide estimates of a single population effect size) are likely to be untenable. Therefore, the random effects model will, in all likelihood, be used for pooling results. Results for randomized experiments and quasi-experimental designs will be pooled and reported separately. As well, results for conceptually-distinct outcomes will be reported separately. If a study reports two separate measures of the same outcome, these will be averaged and the newly created effect size for the same outcome will represent the study in the analysis. Where possible, we will also explore differences between models (e.g., FGC and FUM).
Aron - Odds Ratio Statement
Unit of analysis issues
(a) Cluster-randomised studies
Where trials have used clustered randomization, we anticipate that study investigators would have presented their results after appropriately controlling for clustering effects (robust standard errors or hierarchical linear models). If it is unclear whether a cluster-randomized trial has used appropriate controls for clustering, the study investigators will be contacted for further information. Where appropriate controls were not used, individual participant data will be requested and re-analysed using multilevel models which control for clustering. Following this, effect sizes and standard errors will be meta-analysed in RevMan using the generic inverse method (
(b) Cross-over studies
Due to the nature of the intervention (Family Group Decision Making) we do not anticipate cross-over studies will be identified.
Dealing with missing data
In cases where data are missing (e.g., subgroup means and standard deviations, valid Ns), we will contact the author(s) of the primary studies and try to obtain missing information. We will also search for unpublished reports or other write-ups of the studies. If an author is unable or unwilling to provide information, we will exclude the study if there is inadequate information to proceed. Data on excluded subgroups (e.g., program drop-outs) will be sought as well and Intent to Treat (ITT) analysis will be conducted wherever possible. Studies using ITT or where an ITT analysis can be conducted will be compared with studies where this information is unavailable.
Assessment of heterogeneity
Statistical heterogeneity in the outcome measures will be assessed using the Q-statistic and the associated p-value for each analysis and the I2 statistic (
Assessment of reporting biases
Publication and small sample bias will be assessed with graphical inspection of funnel plots, and “trim and fill” methods that estimate treatment effect by adjusting for the number and outcomes of missing studies (
The authors will deal with selective outcome reporting by searching for the original reports upon which many of the included studies will have been based and comparing the types of outcomes included with the outcomes reported in the published studies. The authors will also search conference abstracts and other sources of gray literature for earlier or additional studies and compare the range of outcomes. In cases where selective outcome reporting is suspected, the study author will be contacted. The assessment of risk of bias due to selective reporting of outcome will be made for each study as a whole.
Data synthesis
Data synthesis will be conducted using RevMan 5 and Comprehensive Meta-analysis 2.0. The determination of independent findings will be completed using the following procedures: First, studies may have included more than one measure of the outcome; therefore, to ensure statistical independence of study findings, each measure of the outcome will be analyzed separately.
We will do separate analyses for absolute vs. relative effects by doing separate analyses for studies that use no-treatment or wait-list controls and for studies that compare two different treatments (Absolute effects will be compared with no treatment, e.g., wait lists; relative effects will be compared to other treatments, e.g., regular child welfare service provision or TAU).
Multiple outcomes for dependent or overlapping samples (i.e., multiple treatments compared against one control group) will be coded separately. We will select only one effect size for inclusion in the meta-analysis based upon conceptual relevance (i.e., FGDM versus another commonly used intervention), sample size, and completeness of information. For studies that include multiple follow-ups, these will be divided into separate intervals (i.e. effects within 6 months, 7-12, 13-24, more than 24 months) and we will do separate meta-analyses for each separate interval.
If there is substantial and unreconcilable heterogeneity with respect to intervention definition, outcomes specification, or measures used, there may come a point at which a meta-analysis becomes untenable (e.g., the population, intervention, measures, or outcomes differ so substantially that combining studies would make little sense), we will conduct a narrative synthesis of the studies. The narrative synthesis would still address each outcome and would detail the studies included, their methodological strengths and weaknesses (as evaluated using CONSORT guidelines), and, based on these, a conservative appraisal of the merit of the intervention will be provided. This is not the preferred course since such a synthesis would not produce any clear recommendations for practice and policy. Nonetheless, it would provide readers with an honest appraisal of systematically gathered studies rather than a more biased literature review, and might prompt further rigorous studies of this commonly used intervention.
Syntheses of higher quality studies are often considered more accurate than syntheses of lesser quality studies. However, there is the possibility that less rigorous designs may produce less biased effect sizes (i.e., they may measure outcomes better or use less biased samples). Moderator analysis (if possible) will be used to control for some of these differences. However, if different trends emerge based on study design (e.g., RCT's v. non-equivalent control group designs), and these can not be controlled for in the analysis, results will be presented separately and the possible reasons for such differences will be discussed.
Subgroup analysis and investigation of heterogeneity
To the greatest extent possible, methodological and clinical heterogeneity among studies will be explored in terms of variations associated with overall study design (experimental and quasi-experimental designs), baselinecharacteristics (i.e., child(ren) in care v. child(ren) in-home)), type of FGDM (e.g., FGC v. FUM), comparison condition (variations in TAU if found), duration of follow-up, and outcome measures (i.e., different measures of a single outcome). If sufficient studies are found, we will also examine effects of interventions with different subpopulations (e.g., type of maltreatment, type of out-of-home care provider). Moderator analysis will be performed using the ANOVA analog (for categorical moderators) and/or meta-regression (for continuous moderators).
Sensitivity analysis
Sensitivity analysis will be performed to assess the robustness of conclusions to quality of data and approaches to analysis (see risk of bias section). Sensitivity analysis will be performed by reanalysis, excluding studies with poor quality indicators (e.g., high attrition, differential attrition, lack of intent-to-treat analysis, lack of controls for baseline differences).
Footnotes
ACKNOWLEDGEMENTS
The following people graciously responded to requests for information: Lisa-Merkel Holguin, American Humane Association; Ted Keys, Oregon Department of Human Services; Angela Rodgers, Portland State University. We would also like to acknowledge the early contribution of Tony Newman, who originally registered this title. Anne-Marie Jørgensen, SFI Campbell for searches of the Nordic data bases.
CONTRIBUTION OF AUTHORS
Aron Shlonsky contributes substantive child welfare content and systematic review expertise. He is the lead author and will guide all facets of the review including write-up, review of articles, and synthesis.
Kate Schumaker contributes substantive content expertise and is responsible for a substantial portion of the write-up, review of studies, and some synthesis
David Crampton contributes substantive North American content expertise, locating English language grey literature, review of studies, and interpretation of results
Charlene Cook and Michael Saini contribute methodological expertise, including meta-analysis
Elisabeth Backe-Hansen contributes searches of the Nordic published and European grey literature, review of studies, as well as substantive European content expertise
Krystyna Kowalski contributes searches of the Nordic literature, hand searching of English journals, and review of studies
DECLARATIONS OF INTEREST
Aron Shlonsky, Katherine Schumaker, Michael Saini, Charlene Cook, Krystyna Kowalski and Elisabeth Backe-Hansen have no declarations of interest to report.
David Crampton currently receives funding from the Annie E. Casey Foundation to support his participation in an evaluation of Team Decisionmaking (TDM) and other child welfare reforms associated with the foundation's Family to Family Initiative. The Casey Foundation promotes TDM. He previously conducted an evaluation of a Family Group Decision Making program in Kent County in Michigan which was supported by the Grand Rapids Community Foundation, W. K. Kellogg Foundation and the United States Children's Bureau. He has written nine articles about FGDM that were published by the American Humane Association (AHA); AHA promotes FGDM. He has authored or co-authored seven articles about FGDM and/or TDM in peer-reviewed publications. He also wrote a book chapter about the use of FGDM for older youth in foster care with Joan Pennell for which they received support from Casey Family Services.
PUBLISHED NOTES
This protocol is co-registered within the Cochrane Collaboration.
