PROTOCOL: Psychosocial, pharmacological and legal interventions for improving the psychosocial outcomes of children with substance misusing parents

Abstract This review aims to first enhance and update existing reviews by comprehensively synthesising the full array of psychosocial, pharmacological and legal interventions that aim to improve the psychosocial outcomes of children with substance misusing parents. Second, the review aims to use network meta‐analysis to integrate and examine the comparative impact of these interventions. Specifically, the review will address the following research questions: (1) What is the comparative impact of psychosocial, pharmacological, and legal interventions for improving the psychosocial outcomes of children with substance misusing parents? (2) Does the impact of interventions vary according to the child developmental period (e.g., infancy, early childhood, adolescence) or the type of (a) outcome measure; (b) substance misuse; (c) practitioner implementing the intervention; or (d) intervention setting? (3) Does the impact of interventions vary by the country of implementation?

. For example, parents with substance misuse issues often have difficulty regulating negative emotional states or experience co-occurrence of mental health disorders (Smithet al., 2012;Whitaker, Orzol, & Kahn, 2006), which can impact their capacity to assess and attend to their child's emotional wellbeing and needs (Borelli, Luthar, & Suchman, 2010;Borelli, West, Decoste, & Suchman, 2012;Siqveland and Moe, 2014) or responsively parent their child according to child developmental needs (Slesnick, Feng, Brakenhoff, & Brigham, 2014;Velez et al., 2004). A history of trauma and childhood adversity is common in parents with substance misuse problems (Hatzis, Dawe, Harnett, & Loxton, 2019) and this, combined with the risk factors above, make sensitive and responsive caregiving challenging (Hatzis, Dawe, Harnett, & Barlow, 2017). Importantly, deficits in parent emotional regulation and the capacity to responsively parent are key predictors of child abuse and maltreatment (Stith et al., 2009).
Global estimates indicate that ∼5-10% of all children are being raised in families with one or more parent who misuses alcohol or other drugs (Dawe et al., 2006;Jääskeläinen, Holmila, Notkola, & Raitasalo, 2016;Manning, Best, Faulkner, & Titherington, 2009;Raninen, Elgán, Sundin, & Ramstedt, 2015;SAMHSA, 2014). This prevalence and the complexities and enduring challenges associated with parental substance misuse has led to the development of a range of approaches that aim to reduce risk factors, enhance family functioning, and improve child outcomes. Importantly, recent estimates suggest that for every dollar invested into substance misuse treatment, there are significant cost savings for society (Dalziel, Dawe, Harnett, & Segal, 2015;National Institute of Drug Abuse, 2012;Public Health England, 2014). However, a critical limitation of the current evaluation and review literature is the lack of integration and synthesis of the relative effectiveness of different intervention models that aim to improve the outcomes for children with substance misusing parents. Without a clear understanding of the relative effectiveness of different intervention approaches, practitioners and policy-makers are limited in their ability to make informed and reliable choices between intervention models. Therefore, the proposed review aims to provide a comprehensive up-todate review of psychosocial, pharmacological, and legal interventions in the context of parental substance misuse and the impact of these interventions on child psychosocial outcomes. Moreover, the review will provide a unique contribution by using network meta-analysis (NMA) to synthesise the comparative effectiveness of these different intervention approaches (see Hutton et al., 2015;Mavridis, Giannatsi, Cipriani, & Salanti, 2015;Salanti, 2012;Wilson, Tanner-Smith, & Mavridis, 2015).

| The intervention
In order to conduct an NMA, this review will include all possible interventions that explicitly aim to improve the psychosocial well-being of families in which at least one parent has either a current substance misuse problem or is in treatment for substance misuse problems. The focus of the review will be on studies that examine the impact of psychosocial, legal and/or pharmacological interventions on child psychosocial outcomes. For the purposes of this review, we draw on Maynard et al. (2015) to define a psychosocial intervention to encompass those that are implemented by professional practitioners (e.g., clinicians, social workers, teachers) across a variety of settings (e.g., homes, school, community, clinics, residential facilities and/or hospitals) that aim to address psychological and social wellbeing more generally. We draw on Eggins et al. (2020) and Mazerolle, Eggins, and Sydes (2018) and define a legal intervention to be "some kind of a strategy, technique, approach, activity, campaign, training, directive, or funding or organisational change that involves the criminal justice system" (p. 21). We define pharmacological interventions to be medication or pharmacy-related approaches to treating parental substance misuse (e.g., buprenorphine or morphine). Examples of eligible interventions are briefly described below (not exhaustive) and additional inclusion specifications are provided in Section 3.

| Home visiting
Home-visiting interventions are characterised by regular home visits made by health practitioners, such as nurses, or paraprofessionals to either pregnant mothers or mothers and their young infants. Homevisiting programmes generally aim to improve the psychosocial and health outcomes for mothers and infants (Segal, Sara Opie, & Dalziel, 2012;Turnbull & Osborn, 2012). This category of interventions generally begins in either the prenatal or early postpartum period and the duration of the intervention can span from weeks to years (Segal et al., 2012). The specific content of home-visiting interventions varies and can include psychoeducation, health surveillance, connection with community resources, parent training and/or counselling components (Turnbull & Osborn, 2012). Home-visiting interventions with substance misusing parents have been evaluated with randomised controlled trials using child psychosocial outcomes US and non-US locations (e.g., Butz et al., 2001;Quinlivan, Box, & Evans, 2003;Schuler, Nair and Kettinger, 2003).

| Family treatment drug courts
Family treatment drug courts (FTDC) are specialised problem-solving courts that use a nonadversarial and treatment-oriented approach for managing cases where both parental substance misuse and child maltreatment have been identified as issues (Gifford, Eldred, Vernerey, & Sloan, 2014;Gifford, Sloan, Eldred, & Evans, 2015;Lloyd, 2015). Cases dealt within FTDCs have a dual focus on promoting the safety and well-being of children and families and also the treatment of parents' substance misuse, whereby withdrawal or retention of parental rights is used as leverage for treatment compliance (Dakof et al., 2010;Gifford et al., 2014;Lloyd, 2015). Although the exact process and content can slightly differ across jurisdictions, common FTDC components include: (a) multidisciplinary teams of professionals who collaborate with families to devise a holistic case-plan to address parental substance misuse and child welfare issues; (b) frequent court hearings and drug testing to monitor treatment adherence and case progress; (c) incentives or rewards for compliance; and (d) sanctions for noncompliance (Chuang, Moore, Barrett, & Young, 2012;Edwards & Ray, 2005;Haack, Alemi, Nemes, & Cohen, 2005). FTDCs have been most widely evaluated in the United States (e.g., Ashford, 2004;Worcel, Furrer, Green, Burrus, & Finigan, 2008), yet have recently emerged in the United Kingdom (Bambrough, Shaw, & Kershaw, 2014) and Australia (Marshall, 2015).

| Multidimensional interventions
In order to address the accumulation of risks across multiple ecological domains characteristic of families with parental substance misuse issues, a number intervention models integrate substance misuse treatment with other biopsychosocial treatments (Marsh, Smith, & Bruni, 2011;Niccols et al., 2012;Uziel-Miller & Lyons, 2000). These multidimensional interventions aim to comprehensively treat parental substance misuse, alleviate other psychosocial risks and minimise barriers to treatment by simultaneously providing intervention components across different ecological domains (Niccols et al., 2012). Common components include: substance misuse treatment (pharmacological and/or psychological support around substance misuse), mental health services, flexible and accessible delivery (e.g., providing transportation and childcare or visiting homes), medical services for family members (e.g., prenatal care, immunisations for children), parenting programmes, vocational and education assistance and other support services (e.g., housing, financial or legal services). Multidimensional interventions have been evaluated or are currently registered for evaluation using randomised controlled or quasi-experimental trials and child psychosocial outcomes measures in a number of countries (e.g., Barlow et al., 2013;Catalano, Gainey, Fleming, Haggerty, & Johnson, 1999;Dawe & Harnett, 2007;Field et al., 1998;Noether et al., 2007).

| Family, parent or child-focused interventions
Interventions within this category can be distinguished from those in the above-mentioned categories based on their narrower intervention focus. Generally, these interventions target the family unit, parents or children in the absence of more intensive casemanagement components. For example, the "Strengthening Families" programme consists of concurrent parent training, child training to promote coping, communication and resistance skills and joint family sessions to facilitate the transfer of acquired knowledge and skills (see Renk et al., 2015 for a review). Other interventions aim to improve psychosocial outcomes of children with substance misusing parents through Behavioural Couples Therapy with parents, sometimes with a parent-training component (e.g., Kelley & Fals-Stewart, 2002). Other interventions in this focused category include pharmacological treatment of the parent's substance misuse (e.g., Coyle et al., 2012;Fischer et al., 1999), school-based psychoeducational programmes (e.g., Dore, Nelson-Zlupko, & Kaufmann, 1999;Gance-Cleveland & Mays, 2008) and attachment-based parenting programmes (e.g., Luthar, Suchman, & Altomare, 2007;Suchman, DeCoste, McMahon, Rounsaville, & Mayes, 2011). In many instances, these interventions are delivered alongside and compared to standard treatment (e.g., methadone maintenance or usual casemanagement practices).

| How the intervention might work
Due to the wide range of interventions that will be included in this review, there are a number of possible mechanisms by which intervention might work. In a general sense, interventions for substance misusing parents are likely to impact child outcomes by modifying or reducing the impact of known risk factors, including parental psychopathology, parenting knowledge and skills, enhancement of the quality of the parent-child relationship, involvement in the criminal justice or child welfare systems or impoverished environments. Different categories of interventions may impact child outcomes through more specific pathways. For example, interventions based on family disease models generally use a 12-step model, focusing on abstinence, psychoeducation and knowledge as key mechanisms through which change can occur (for a review, see Usher, McShane, & Dwyer, 2015). In comparison, family prevention models target risk and protective factors linked with parent substance misuse to generate change (for a review, see Usher et al., 2015). Pharmacological interventions reduce ingestion of drugs, thereby reducing foetal exposure and improving birth and developmental outcomes (Minozzi, Amato, & Vecchi, 2013). Legal interventions, such as Family Treatment Drug Courts, aim to reduce the risks associated with parental substance misuse through providing support, whilst motivating behaviour change through incentives and penalties (Fay & Eggins, 2019). Therapeutic or psychosocial interventions also aim to improve child outcomes by reducing the risks associated with parental substance misuse, yet the theoretical models that underpin interventions in this category vary depending on the focus of the intervention (for reviews, see Dawe & Harnett, 2013;Neger & Prinz, 2015).

| Why it is important to do the review
The current evaluation and review literature lacks integration and synthesis of the relative or comparative effectiveness of different intervention models that aim to improve the outcomes for children with substance misusing parents. Yet without a comprehensive and integrated synthesis of the extant evaluation literature, it will be difficult EGGINS ET AL. | 3 of 20 for practitioners, policy-makers and researchers to focus their resources and decision-making to improve the lives of children and families with substance misusing parents. As of December 2018, there were 24 existing reviews that (a) focus on interventions specifically for substance misusing parents and (b) have captured one or more studies that have assessed the impact of an intervention on child psychosocial outcomes. Although not all of these reviews adhere to a full systematic review methodology, each employs at least two systematic review techniques (e.g., systematic search, specific inclusion criteria and qualitative or quantitative synthesis of studies) and can be considered less biased than narrative reviews in the area (e.g., see, Choi, 2012;Marsh et al., 2011;Oliveros & Kaufman, 2011;Renk et al., 2015).
These reviews highlight the range of interventions and large number of studies that are necessary for conducting an NMA. Existing reviews differ according to the specific intervention under consideration and whether only child outcomes or multiple different types of outcomes are included, and can be summarised as follows.
• One review examines the impact of home-visiting interventions during pregnancy and the postnatal period for women with substance misuse issues and their impact across a range of parental and child outcomes (Turnbull & Osborn, 2012).
• Three reviews focus on Family Treatment Drug Courts for substance misusing parents with and their impact on child out-ofhome placement (Llyod, 2015), child welfare system outcomes (Zhang, Huang, Wu, Li, & Liu, 2019) or child maltreatment outcomes (Eldred & Gifford, 2016).
• One review examines the impact of multidimensional interventions for substance misusing mothers and their impact on multiple child outcomes (Niccols et al., 2012).
• Two reviews and one Cochrane protocol concentrate on childfocused preventative interventions for improving outcomes for children of substance misusing parents (Bröning et al., 2012) or alcohol misusing parents (Cuijpers, 2005 • Two Cochrane reviews examine the impact of pharmacological interventions during pregnancy on maternal and child outcomes in the context of alcohol misuse (Smith, Lui, & Terplan, 2009) and opioid dependence (Minozzi et al., 2013).
• Several reviews capture a broad range of psychosocial interventions for parental substance misuse and their impact on multiple outcomes (including child outcomes) for either alcohol misuse during pregnancy (Lui, Terplan, & Smith, 2008;Scobie and Woodman, 2017;Stade et al., 2009)  Although the existing review literature is extensive, there is variation in the degree of methodological quality and content coverage. We argue that issues pertaining to the methodologies used for existing reviews and gaps in content coverage reduce the ability to draw reliable conclusions about the effectiveness of interventions for improving psychosocial outcomes for children with substance misusing parents.

| Methodological limitations of existing reviews
Perhaps the most important methodological limitation of existing reviews is the lack of quantitative syntheses. Very few of the reviews with sufficient studies use meta-analysis to synthesise the evaluation evidence (e.g., Minozzi et al., 2013;Niccols et al., 2012;Turnbull & Osborn, 2012;Zhang et al., 2019), despite the availability of multiple studies suitable for meta-analysis across many of the reviews. Rather, authors provide qualitative summaries of intervention effectiveness that are based on the raw differences, statistical significance or effect sizes of individual studies. Although qualitative summaries are useful for assessing the breadth and qualities of intervention research, this methodology is inadequate for providing a reliable and precise estimate of an intervention impact (Borenstein, Hedges, Julian, & Rothstein, 2009;Littell, Corcoran, & Pillai, 2008).
Methodological limitations of existing reviews also highlight the need for an updated and more comprehensive systematic search.
First, existing reviews may not provide an accurate representation of the most up-to-date intervention evidence because between 5 and 10 years have passed since the searches were conducted for many of the reviews. Second, there may be potential biases in the existing reviews. Some authors excluded studies that found negative intervention effects or only reported study outcomes if they were statistically significant. Others have introduced publication bias by either explicitly excluding documents not published in peer-reviewed journals, omitting searches for unpublished literature or limiting searches to very few sources. In addition, some authors implemented restrictive searches, such as very limited search terms or multiple Boolean AND clauses. Third, much of the current body of reviews lacks transparency in the reporting of searches and sensitive search strategies. Many authors do not explicitly report their exact search and how it was implemented during their systematic search (e.g., what search fields were used).

| Content gaps in existing reviews
The current corpus of reviews does not provide complete coverage of the extant literature. Some reviews explicitly omit studies that include substance misusing fathers (e.g., Niccols et al., 2012), focus only on the prenatal period (e.g., Minozzi et al., 2013;Smith et al., 2009) or omit studies that contain child outcomes in the absence of parentlevel outcomes (e.g., Neger & Prinz, 2015). Others focus on alcohol misuse and do not capture equivalent interventions for populations with illicit drug misuse issues (e.g., Cuijpers, 2005;Lui et al., 2008;Smith et al., 2009;Stade et al., 2009;Templeton et al., 2010).
However, the most important limitation is that the existing review literature does not permit valid conclusions to be made about the comparative impact of these interventions for children with substance misusing parents. Yet understanding the relative impact of different interventions for a particular population is crucial for informing the decision-making of both practitioners and policy-makers (Hutton et al., 2015;Mavridis et al., 2015;Salanti, 2012). A recent methodological development, called NMA, provides an avenue for addressing these this important question. NMA, also known as multiple treatments meta-analysis, has been referred to as "the next generation evidence synthesis tool" (Salanti, 2012, p. 80) and extends traditional pairwise meta-analytic techniques. NMA provides an approach for (a) quantitatively synthesising both direct and indirect effects of multiple interventions for a particular population or condition; and (b) ranking interventions according to their effectiveness, even in the absence of trials that have directly compared the treatments Salanti, 2012).
Our review will begin to address the above-mentioned methodological quality and content coverage issues. Specifically, our review will both (a) enhance and update the existing body of reviews and (b) synthesise the comparative impact of interventions on the psychosocial outcomes of children with substance misusing parents. Provided sufficient data are available and the underlying analytical assumptions are satisfied, the proposed review will provide the first NMA that synthesises the relative impact of multiple interventions on the psychosocial outcomes for children with substance misusing parents. Importantly, the review will enable policy makers and practitioners to make informed and reliable choices between intervention models.

| OBJECTIVES
The overarching objective of this review is twofold. First, we aim to enhance and update existing reviews by comprehensively synthesising the full array of psychosocial, pharmacological and legal interventions that aim to improve the psychosocial outcomes of children with substance misusing parents. Second, we aim to use NMA to integrate and examine the comparative impact of these interventions. Specifically, the review will address the following research questions.
• What is the comparative impact of psychosocial, pharmacological and legal interventions for improving the psychosocial outcomes of children with substance misusing parents?
• Does the impact of interventions vary according to the child developmental period (e.g., infancy, early childhood, adolescence) or the type of (a) outcome measure; (b) substance misuse; (c) practitioner implementing the intervention; or (d) intervention setting?
• Does the impact of interventions vary by the country of implementation? (b) the likelihood that risk of bias will affect specific design features versus an overall research design category. For the purposes of this review, methodologically robust quasi-experimental designs are defined as those which permit causal inference by minimising threats to internal validity. For example, maximising treatment and comparison group equivalence through matching (e.g., propensity score matching), measurement of outcomes multiple times pre-and postintervention to reduce maturation threats (e.g., interrupted timeseries, cohort panel designs) or adjusting for confounding factors through statistical modelling (e.g., multiple regression, propensity score modelling). Due to serious threats to internal validity, single group studies with one preintervention and one postintervention outcome measure will be excluded from the review.
To be included in the meta-analyses, study authors must report sufficient data to calculate an effect size. Where data are not reported, the required data will be sought by contacting the document authors. Should the data not be provided by study authors, the study will be excluded from meta-analyses, but will be coded and included in study summary tables. the care of at least one a parent, and a family is defined as at least one child and one parent.

| Types of participants
Parents will be classified as "currently substance misusing" if they have been classified as such via standardised diagnostic criteria (e.g., DSM, ICD 10) or a self-report measure (e.g., AUDIT). In the absence of classification supported by diagnostic or self-report measures, studies will be included if the authors explicitly identify the research population as substance misusing parents. For example, a study would be included if the authors note that all study participants are methadone maintained mothers, even if the authors do not report formal diagnoses or baseline levels of substance use. Parents will be classified as substance misusing if they are misusing alcohol, illicit drugs and/or prescription drugs. If the study sample is not comprised completely of substance misusing parents, we will follow Turnbull and Osborn's (2012) approach, whereby the study sample must include at least 50% substance misusing parents to be included in the review.

| Types of interventions
In order to conduct an NMA, this review will include all possible psychosocial, pharmacological or legal interventions that explicitly aim to improve the psychosocial well-being of families characterised by parental substance misuse. However, the focus of the review will be studies that examine the impact of interventions on child psychosocial outcomes. Examples of eligible interventions are described in the Section 1.2 (not exhaustive) and decisions regarding the categorisation of interventions will be based on the TIDieR Checklist (Hoffmann et al., 2014) and the specific intervention components coded for each eligible study (see Appendix). The categorisation approach will be reported in the final review and will guide assessment of the transivity assumption (Hutton et al., 2015). Based on existing literature in the area, we anticipate that the majority of the included studies will utilise a treatment-as-usual comparison condition (e.g., methadone maintenance, case-management without the intervention under consideration). Interventions will be included irrespective of whether it is initiated during the prenatal or postnatal period. In addition, studies will be included if the intervention focuses on the misuse of alcohol, illicit drugs and/or prescription drugs.

| Types of outcome measures
In order to comprehensively synthesise the impact of eligible interventions on children with substance misusing parents, this review will include a broad range of outcomes nested under the banner of "psychosocial wellbeing". Outcomes will be considered eligible if they are measured using standardised or nonstandardised instruments or consist of official, diagnostic, observation or self-report data.
Examples of primary outcomes include, but are not limited to, the following.
The decision to utilise one intervention over another may rest on other considerations beyond the effectiveness of the intervention, such as intervention cost, resource intensity or degree to which participants accept or complete treatment. Therefore, if reported in eligible studies, the following secondary outcomes will also be coded and analysed: cost-effectiveness, treatment completion, length of time in treatment and acceptability of treatment (e.g., participant perspectives of the intervention).

| Duration of follow-up
Studies will be included irrespective of the length of follow-up after the intervention. Where the length of follow-up varies across studies, we will group and synthesise studies according to similar follow-up durations. For example, short (e.g., 0-3 months postintervention), medium (>3 months, <6 months) and long-term follow-up (>6 months postintervention).

| Types of settings
There will be no restrictions on the intervention setting or treatment format (e.g., inpatient, outpatient, community settings, family home, online or computerised, one-on-one or group settings). The review will include intervention studies conducted in any geographical location or country and published in any language, provided translation can be undertaken to code eligible studies.

| Additional eligibility criteria
Studies that satisfy the above-mentioned eligibility criteria will be included in the review regardless of publication status. While documents written in languages other than English will not be excluded from the review, they will only be assessed for final eligibility and included in the syntheses if a translation can be sourced.
Due to differences in the search operators and indexing systems across electronic databases, the exact search will be customised to each location and may slightly vary across search locations. Where the functionality of a search location does not permit complex search strategies, a simplified version of the search will be utilised. The search will place no limits on publication date, document language or (e.g., book reviews). Each search will be recorded in a search record as per recommended guidelines (see Kugley et al., 2017;Littell et al., 2008).
To reduce disciplinary and publication bias, the systematic search will cover multiple disciplines and search sources (see Table 2 for search locations). The following additional search strategies will also be used to identify eligible documents not already captured.  Table 2).

| Description of methods used in primary research
Based on the existing corpus of reviews, we anticipate that included studies will employ either randomised or nonrandomised experimental designs with a range of comparison conditions (e.g., treatment-as-usual, alternative treatment, waitlist control). In addition, we anticipate identifying studies that assess intervention impact using a broad range of child psychosocial outcomes measured in a continuous and/or categorical manner.

| Criteria for determination of independent findings
There are three areas of consideration for determining independence of study findings: multiple reports of the same study, multiple outcomes for the one study and research designs with clustering. The software that will be used to manage this review enables nesting of multiple dependent documents pertaining to one study. If dependent studies are identified, all studies will be coded and data will be extracted from the most complete report of the study and the study will only be included once in each meta-analysis for each conceptually unique outcome category. Where studies report on multiple conceptually similar outcomes, the effect sizes will be combined into a single effect size using the method described by Borenstein et al. (2009). This composite effect size will then be used in the metaanalysis. If a study utilises a research designs with clustering (e.g., study sites assigned to conditions), the method suggested by Fu et al. (2013) and Higgins et al. (2011) will be used to adjust the standard error. If the included study does not report the required intraclass correlation coefficient (ICC), we will use the approach taken by Barlow, Bergman, Kornør, Wei, & Bennett (2016)

| Title and abstract screening
The first phase of assessing study eligibility will be comprised of screening the titles and abstracts identified in the systematic search.
After removing duplicates and clearly ineligible document types (e.g., book reviews), records captured by the systematic search will be imported into the SysReview review management software (Higginson & Neville, 2014). Each title and abstract (record) will then be assessed according to the following exclusion criteria.
2. Document is not unique.

Document is not about parental substance misuse.
Although all efforts will be made to remove ineligible document types and duplicates prior to screening, automated and manual cleaning can be less than perfect. Therefore, the first two exclusion criteria will be used to remove ineligible document types and duplicates prior to screening each record on substantive content relevance.
The full-text document for each record retained at the title and abstract screening stage will then be attached within SysReview before progressing to full-text eligibility screening. If full-text documents cannot be retrieved via existing university resources, they will be ordered through the university libraries of the review authors or by contacting study authors.
2. Document is not unique.

Not a quantitative impact evaluation of an eligible intervention
using eligible participants or outcomes.

Ineligible research design.
Although all efforts will be made to remove ineligible document types and duplicates in prior stages, these types of records can sometimes progress into later stages, for example, where duplicate records are not adjacent to each other during screening or where screeners cannot unequivocally determine if record is ineligible based on the title and abstract. Therefore, the first two exclusion criteria will be used to remove ineligible document types and duplicates prior to screening each document for final eligibility.

| Full-text coding and risk of bias assessment
Eligible studies progressing from the full-text screening stage will be independently double-coded using the coding protocol provided in the Appendix. Broadly, studies will be coded according to the following domains.

Risk of bias.
Risk of bias will be assessed using either the Cochrane randomised or nonrandomised risk of bias tools Sterne et al., 2016), whereby studies will be rated across domains as having high, low or unclear risk of bias. If a domain is rated as "unclear", study authors will be contacted to obtain missing data.
Results of the risk of bias assessment will be presented in summary tables and in a risk of bias summary figure. Depending on the data available, sensitivity analysis will be used to examine the impact of risk of bias on effect estimates and corresponding con-
We anticipate that primary studies will report a variety of data for calculation of effect sizes. Where outcomes are reported as continuous measures, Hedges g (standardised mean differences [SMDs]) will be computed, along with 95% confidence intervals. For studies with binary outcomes, effect sizes will be computed as odds ratios and then transformed into SMDs for meta-analyses (see Borenstein, Hedges, Higgins, & Rothstein, 2009). For all other studies not falling into these categories, we will use David B. Wilson's effect size calculator to calculate SMDs for meta-analyses.
We will follow the recommendations of the Campbell Collaboration Methods Group and PRISMA Group when conducting and reporting on networked statistical analyses (Hutton et al., 2015;Wilson, Tanner-Smith, & Mavridis, 2016). Specifically, the geometry of the network(s) will be presented in network diagrams, whereby nodes represent competing interventions and/or comparisons and the lines connecting the nodes reflect direct comparisons between nodes (Chaimani, Higgins, Mavridis, Spyridonos, & Salanti, 2013;Mavridis et al., 2015). The weight of the lines in network diagrams will represent the number of studies with the respective comparison and the size of the nodes will reflect the number of studies involving the respective intervention. The transivity assumption will be assessed by examining the consistency between the direct and indirect evidence, drawing on a global test approach (see Hutton et al., 2015 for a review) and inconsistency plots (see Chaimani et al., 2013;Mavridis et al., 2015). The relative effects of each pair of interventions will be summarised in league tables, and intervention rankings will be displayed using rankograms and ranking plots, with accompanying surface under the cumulative ranking curve values (see Hutton et al., 2015;Mavridis et al., 2015). Where there is inconsistency detected, intervention rankings will be interpreted with caution.
It is important to note that NMA assumes that there are effect sizes that compare (a) eligible treatments and comparison groups; and (b) eligible treatments with other eligible treatments. If there is an absence of the latter, it may be more appropriate to conduct moderator analyses used in standard meta-analyses, which are conducted using either the analogue to the analysis of variance (ANOVA) or metaregression approach. Both approaches require sufficient studies to reach statistical power, but the metaregression approach will be the preferred approach due to the ability use multiple variables which may impact effect size estimates (e.g., design features, risk of bias, focus of treatment).
Due to the difficulties that can be associated with meeting the transivity assumption, consistency assumption and types of within study comparisons, it is important to specify an alternative a priori analysis approach. One part of the problem stems from only a preliminary understanding the extant evaluation literature that may be eligible for the review when finalising a review protocol. For example, the systematic review may identify nuances in the interventions and outcome categories that require an approach that was not anticipated at the protocol stage. We anticipate conducting separate network meta-analyses for each conceptually distinct outcome category that will combine all comparison conditions. However, should this not be possible, we will conduct traditional meta-analyses that examine each broad intervention model as compared to (a) eligible comparison conditions; and (b) alternative treatments. To categorise interventions, we will use the TIDieR Checklist (Hoffmann et al., 2014), along with the intervention components using the coding guide in the Appendix, to ensure that intervention categories are internally consistent.
For studies reporting multiple points of follow-up, effect sizes will be calculated for each time-point, but synthesised separately with studies that have similar outcome time-points. If included studies report baseline and postintervention outcome data, SMDs will be calculated using baseline adjusted mean differences (i.e., mean change scores) and the change score standard deviations, will be standardised using the raw standard deviation within groups. Where authors do not report the standard deviation for mean change scores, Lipsey and Wilson's (2001)  . If studies report follow-up outcome data, postonly outcome data will be used to estimate SMDs, and follow-up outcomes will be analysed separately from postintervention outcomes.
Heterogeneity between studies will be examined using the approach suggested by Chaimani et al. (2013). Specifically, τ 2 will be used to express heterogeneity across the network, and we will present the NMA mean summary effects in forest plots with corresponding confidence and predictive intervals to assess whether the magnitude of heterogeneity impacts the mean summary effects.
Moderator analyses will be used to explore potential sources of heterogeneity. Specifically, the analogue to ANOVA will be used for categorical moderators and regression-based approaches will be used for continuous moderators. Depending on the data reported in included studies, additional exploratory subgroup analyses may be performed; however, we will clearly distinguish between a priori and exploratory analyses in our reporting. The specific a prior moderator variables are: child developmental period (e.g., infancy, early childhood, adolescence); type of outcome measure (e.g., mental health, child welfare placement, child abuse or neglect); type of parental substance misuse (e.g., alcohol, drug, polysubstance); type of practitioner implementing the intervention; type of intervention setting (e.g., home, school, community, residential facility); and country of implementation.
Assessment of publication bias will be the final stage of analysis and will first entail inspection of comparison-adjusted funnel plots for asymmetry to identify whether effect size estimates are influenced by publication bias (see Chaimani et al., 2013;Mavridis et al., 2015).
If asymmetry is detected, subgroup analyses will be conducted to assess whether effect sizes significantly differ by publication status of the included studies. Effect size data* *To be completed for each eligible outcome within a study (or group of reports for a study). To add another outcome, click the "Add another outcome" button located at the bottom of the screen.