Evidence for investing in parenting interventions aiming to improve child health: a systematic review of economic evaluations

A comprehensive review of the economic evidence on parenting interventions targeting different aspects of child health is lacking to support decision-making. The aim of this review is to provide an up to date synthesis of the available health economic evidence for parenting interventions aiming to improve child health. A systematic review was conducted with articles identified through Econlit, Medline, PsychINFO, and ERIC databases. Only full economic evaluations comparing two or more options, considering both costs and outcomes were included. We assessed the quality of the studies using the Drummond checklist. We identified 44 studies of varying quality that met inclusion criteria; 22 targeting externalizing behaviors, five targeting internalizing problems, and five targeting other mental health problems including autism and alcohol abuse. The remaining studies targeted child abuse (n = 5), obesity (n = 3), and general health (n = 4). Studies varied considerably and many suffered from methodological limitations, such as limited costing perspectives, challenges with outcome measurement and short-time horizons. Parenting interventions showed good value for money in particular for preventing child externalizing and internalizing behaviors. For the prevention of child abuse, some programs had the potential of being cost-saving over the longer-term. Interventions were not cost-effective for the treatment of autism and obesity. Future research should include a broader spectrum of societal costs and quality-of-life impacts on both children and their caregivers. Supplementary Information The online version contains supplementary material available at 10.1007/s00787-022-01969-w.


Introduction
There is a widespread acceptance that the interplay between the child and her environment, composed of many layers including family, peers, and social structures, contribute to the child's development.Of all protective and risk factors influencing child development, a key part is the quality of parenting that children are exposed to [1].Inadequate parenting, such as low levels of supervision and involvement, inconsistent rule setting, and punitive discipline, therefore, reinforce inappropriate or negative outcomes in children [2].Conversely, a warm and close relationship where parents are supportive and use positive reinforcement are important protective factors [3,4].
The use of parents as active agents in influencing an unwanted behavior or outcome in children may thus be beneficial, an idea that has been formative in the development of parenting interventions.Parenting interventions aim to improve child behavior through improving parenting strategies and parent-child relationships [5].These are commonly used as a preventive or treatment measure and are largely effective strategies, where changes in parental behavior trickle down to reducing child problematic behavior.This is especially true for disruptive behavior problems [6] and child maltreatment [7].Effects increase per level (universal, selective, and indicated) of prevention [8], and parenting interventions are effective when delivered face-to-face [9] as well as online [10].Effects have also been seen across different country contexts [11].
Problems that parenting interventions aim to target, including disruptive behavior, child maltreatment, emotional problems, and obesity, are costly to individuals, families, and society [12,13].If coexisting with parental mental health problems, they may be even larger [14].If problems are left unresolved, they may additionally result in long-term adverse consequences, including persistent mental health problems, socio-economic struggles, and criminality [15][16][17].Simultaneously, to provide the best possible and equitable care to children and their families, resources need to be prioritized in light of competing alternatives.It is, thus, important that decisions on which parenting interventions are to be adopted are made not only based on effectiveness but also on whether the outcomes produced by such interventions are good value-for-money.
The literature on the economics of parenting interventions for improving child health dates back to the 1980s.Considerable research has been undertaken, looking at outcomes and/or costs of parenting interventions and their impact on patterns of resource use.Although informative, evaluations that investigate either outcomes or costs separately, only consider one of two important dimensions of economic evaluation, and cannot fully guide decisionmakers in resource prioritization.Full economic evaluations are necessary, to compare two or more interventions in terms of costs and health outcomes [18].
Several reviews of economic evaluations of parenting interventions have previously been conducted.An earlier review compiled the economic evidence of parenting interventions that aimed to support families with children with or at risk of developing conduct disorder.The review included six studies but was cautionary to draw conclusions; as only three of these studies were full economic evaluations and measured different health outcomes and costs of different interventions, making comparisons difficult [19].Duncan et al. [20] synthesized the economic evidence for parenting interventions aimed to improve parent-infant interaction.Ten studies were included in the review that concluded that the interventions could result in substantial savings, both in the short-and longterm.Another review evaluated the evidence for parenting interventions, in the UK, for preventing behavior problems in children, finding that the interventions had the potential to be cost-saving in the long-term [21].
Lacking in the literature is a comprehensive review including a broader range of child health outcomes, as well as an assessment of study quality.In addition, there has been a surge of economic evaluations of parenting programs in the last five years, more than doubling the amount of available evidence.The aim of the current systematic review is to provide an up-to-date synthesis and appraise the quality of the available health economic evidence for parenting interventions aiming to improve child health.

Search strategy
This systematic review was guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines [22].An English-language literature search was undertaken in Medline, Econlit, ERIC, and PsychInfo for peer-reviewed literature published until June 2020.Search terms were informed by previous systematic reviews [19,21] and included 'child*' AND 'parent*' AND 'economic evaluation' OR 'cost benefit' OR 'cost effectiveness' OR 'cost utility' OR 'cost offset' OR 'cost minimization'.All results were exported to Mendeley version 1.19.4.The review was registered in PROSPERO: CRD42020206303.
Inclusion criteria for the review comprised: (1) studies that met the criteria for a full economic evaluation, considering both costs and outcomes of two or more alternatives; and (2) studies with a randomized or quasi-randomized controlled design with at least one parenting intervention arm aiming to improve child health-related outcomes, including measures of physical and mental health.A parenting intervention was defined as a manualized structured intervention, focusing on parenting skills and practices [5].The following exclusion criteria were considered: (1) reviews, editorials and abstracts from conferences, (2) studies without access to full text; (3) studies where only parental outcomes and no child health-related outcomes were reported.To validate the search strategy and to find missing articles, reference lists of all systematic reviews identified in the initial search were checked for other relevant articles.
Three of the authors (FS, CN, and IF) screened one third of the titles and abstracts each.To consider inconsistencies between author assessments, a random sample of 20% of the abstracts was assessed by another author.Author agreement on article relevance was estimated based on inter-rater reliability, resulting in a Cohen's kappa coefficient between 0.88 and 1.00, reflecting good agreement [23].Inconsistencies were discussed to reach complete agreement.Thereafter, abstracts included were screened for full text inclusion based on inclusion criteria.

Quality assessment
Three independent reviewers (FS, CN, and IF) assessed the quality of the studies included after full text screening using the Drummond [18] checklist.Disagreements were discussed among authors until consensus was reached.This checklist includes ten items, each with three potential responses, "yes", "unclear", and "no", which were scored 1, 0.5, and 0, respectively.Items 7 and 8 have the additional potential response "not applicable".When these items were deemed not applicable, they were excluded from the calculation of total score.A scoring system was used to calculate an average score across all applicable items, with each item weighted equally [24].Total scores range between 0 and 1.Studies were classified into high (score 0.8-1.0),moderate (score 0.6-0.79),and poor quality (score ≤ 0.59).

Data extraction and study classification
Data from the articles selected for inclusion were extracted using a template relevant to the study aim.Extracted items were summarized in a narrative format and included: author/year, setting, problems targeted, population, intervention(s), comparator, follow-up/time horizon, analysis perspective, costs included, outcomes (generic and clinical, and instrument), summary of results, and study quality.Studies were classified according to the type of evaluation performed.Evaluations included cost-effectiveness analyses (CEA) (using clinical outcome measures), cost-utility analyses (CUA) (using generic outcome measures, such as quality-adjusted life-years (QALY) or disability adjusted life years (DALY), which serve as common metrics that can be used to compare different interventions), cost-benefit analyses (CBA) (quantifying health outcomes in monetary terms), cost-minimization analyses (CMA) (when health outcomes are not significantly different, thus only costs are compared), and costconsequence analyses (CCA) (when both costs and outcomes are described without incremental estimates being computed).CUA is the only evaluation type that allows for results to be compared against an established willingnessto-pay (WTP) threshold for a gained QALY.Although it could be argued that cost-offset analyses (comparing costs incurred with costs saved) are not full evaluations, these were also included, as the line between costs and outcomes if often ambiguous and some outcomes might be proxied by service use.Studies were also classified as to whether they were based on primary data or simulation modeling.Interventions were classified within the prevention spectrum (universal, selective, indicated) or as treatment.Universal interventions target the whole population; selective interventions target at-risk groups; and indicated interventions target high-risk groups with signs or symptoms of disorders, but who do not meet the full criteria for a diagnosis [25].
Three reviewers (FS, CN, and IF) independently extracted data, while a random sample of 20% of the articles were selected for review by another author.Disagreements were resolved through discussions.

Search results
The electronic search resulted in 1714 unique publications, and four additional articles were identified when screening reference lists of relevant reviews.Figure 1 shows a flowchart of the study selection process.Based on the title and abstract review, 86 articles were selected for full text review.Of these, 42 were excluded because they did not evaluate parenting interventions (n = 15), did not include a comparator (n = 6), were not full economic evaluations and reported only costs (n = 4) or only effects (n = 4), did not have a relevant outcome (n = 3), did not set costs in relation to effects (n = 3), and were report versions of other publications or not available in full text.After the full text review, 44 studies were selected for data extraction and quality assessment.

Quality assessment
Of the 44 studies included, a majority of the studies were rated to be of high quality (n = 32), 11 studies of moderate quality and 1 of poor quality.The most frequent reasons for not receiving full points among all studies was failing to include all relevant costs and outcomes (64% of studies), failing to discount costs and outcomes occurring in the future to present value (65% of studies), and failing to adequately characterize uncertainty around the cost and effect estimates (64% of studies).Additionally, moderate quality studies also often failed to pose a well-defined research question, to adequately measure and value costs and outcomes, and to address all issues of concern in the discussion (See Tables S1 and S2 in the Supplementary Appendix).

Overview of the studies
Of the 44 included studies, most targeted child mental health (n = 32): 22 studies targeted externalizing behavior problems (such as attention deficit/hyperactivity disorder and conduct disorder) , five studies targeted internalizing problems (i.e., anxiety) [48][49][50][51][52], and five studies targeted other mental health problems including, for instance, autism, and alcohol abuse [53][54][55][56][57].The remaining studies targeted child abuse (n = 5) [58][59][60][61][62], obesity (n = 3) [63][64][65], and general health (n = 4) [66][67][68][69].Most interventions (n = 30) evaluated preventive strategies, whereas 14 studies evaluated treatment strategies.Most studies were conducted in Europe (n = 24) including the UK, Ireland, Sweden, and the Netherlands, followed by North America (n = 13) and Australia (n = 5).The majority of studies were CEA (n = 28) using clinical outcome measures, nine were CUA using QALYs or DALYs, four were CMA, three cost-offset analyses and one cost consequence analysis.Most studies were within trial evaluations (n = 34), a large proportion with time horizons of 1 year or less (n = 23).Ten studies modeled costs and benefits over longer time horizons, ranging from one year to lifetime.A variety of costing perspectives were employed, ranging from limited to fuller societal perspectives (n = 28) and to narrower third-party payer perspectives (n = 8) sometimes limited to only intervention costs.The main characteristics of the studies are summarized in Table 1, and methods and results are summarized in Table 2.
All costs were converted to 2020 US$ using purchasing power parities.

Mental health
Externalizing behavior problems Among the high-quality studies (n = 14), there was evidence that preventive interventions targeting children with symptoms and parenting interventions delivered as treatment, including group and individual face-to-face programs [26,   27,31,35,43,46] were cost-effective and even cost-saving for targeting externalizing behaviors.CBA analyses of some of these programs targeting prevention showed cost-benefit ratios between US$6.48 and US$17.18 per dollar invested over the long-term [27], with savings to society over a 25 year horizon of $28,994 per family [31], or $13,364 over 20 years per child [46].Another study estimated net benefits on the population level of $31.3 million if a minimum reduction of 25% of cases of conduct problems were achieved [35].One study found incremental cost-effectiveness ratios (ICER) of $778 per DALY averted for group therapy and $15,744 per DALY averted for individual therapy (probability of cost-effectiveness between 99.2% and 99.5%) [43]; and another study reported ICERs between $6,527 and $9,923 per QALY for a parent only intervention compared to a parent and teacher variant [47].Well-established and disseminated parenting interventions, such as the Incredible Years [26,27], and the Triple P-Positive Parenting Program [35,43] were likely to be cost-effective at local WTP thresholds.Among the moderate quality studies (n = 7), group-based interventions had similar outcomes at lower costs compared to individual formats [41], and interventions targeting different combinations of parent, teacher, and child formats yielded better outcomes at higher costs than care as usual [33,40,42,70].These studies were in its majority CEA using clinical outcome measures with no existing WTP values to benchmark their results against, hence no conclusions on value for money can be drawn.

Internalizing behavior problems
All five studies targeting internalizing behaviors were high quality.We found evidence that parenting interventions for the prevention of anxiety were cost-effective and parenting interventions delivered as treatment produced similar outcomes at lower or equal costs in relation to comparators.For example, a group-based preventive intervention, was costeffective with an ICER of $6144 per DALY averted and 99% probability of cost-effectiveness [50].Another study reported a probability of cost-effectiveness of 96% at a WTP between $28,694 and $36,279 for a parent-only intervention versus a parent and child intervention [49].

Other mental health problems
Four studies were deemed high quality.Interventions were cost-effective for the prevention of other mental health problems [56,57] or generated better outcomes at equal costs [53].For example, a selective intervention for divorced mothers generated long-term cost-offsets over 15 years of $1,336 per family [56].An evaluation of two universal interventions for the prevention of alcohol abuse among youth had the potential to delay abuse onset, with benefit ratios between $9.97 and $16.35 per dollar invested [57].No interventions were cost-effective for treating autism [55].

Child abuse and neglect
Among the high quality studies, there was conflicting evidence about the cost-effectiveness of parenting interventions for the prevention of child abuse.For example, one study [59] found that both a home-visiting program and centers providing early education in schools and services for lowincome families were likely to be cost-saving over a lifetime horizon, with a benefit-cost ratio of $7.18 per dollar invested.Another study [60] estimated a net present value saving of $2.4 million for treating 100 families in terms of cases of maltreatment prevented.Conversely, one study found the same home-visiting program not cost-effective at local WTP thresholds, with a probability of cost-effectiveness of 26.7% [58].These two studies differed in terms of the costs included in the analysis, where the former considered a broader range of costs than the latter, including for instance productivity losses.

Obesity
Good quality studies did not support the cost-effectiveness of parenting interventions targeting obesity or reported better outcomes at higher costs than comparators.For example, a family-based community program for parents and children, addressing parenting, lifestyle, social and emotional development was not cost-effective compared to TAU (ICER of $893,536 per QALY and 40% probability of cost-effectiveness [64].Another study reported that a family-based behavioral treatment improved BMI and cost more than an information control [63], although cost-effectiveness cannot be inferred.A moderate quality study showed that mixed group and individual family-based treatment was cheaper than individual only treatment [65].

General health
Evidence on the cost-effectiveness of parenting interventions targeting general health was conflicting.One good quality study reported that a population-based program integrated within care as usual targeting mothers and their children yielded higher effects and was cost-saving compared to TAU [67].Another good quality study showed that a groupbased parenting intervention was not cost-effective at local WTP thresholds (probability of cost-effectiveness of 47% at 5 years and 57% at 10 years).[69].Two moderate quality studies did not support the cost-effectiveness of interventions [66,68].

Discussion
This review aimed to provide an up-to-date synthesis of the available health economic evidence for parenting interventions aiming to improve child health.In the last three decades, 44 studies on the economic value of parenting interventions, that met the inclusion criteria for this review, were published.Most of the studies targeted child mental health (n = 32), in particular externalizing behavior problems, followed by internalizing problems, and other mental health problems.The remaining studies targeted child abuse, obesity, and general health.Seventy percent of studies evaluated preventive interventions.
Pleasingly, most studies were of high (n = 32) to moderate quality (n = 11).Among the studies deemed high quality, parenting interventions showed good value for money, in particular for preventing child externalizing and internalizing behaviors.High-quality evaluations of widely used parenting interventions, such as the Incredible Years and the Triple P, show that they, either: (a) were cost-effective at local WTP thresholds; or, (b) could be cost-saving over the long-term.For the prevention of child abuse, some home-visiting programs had the potential of being cost-saving over a lifetime horizon.Family-based community programs targeting the treatment of obesity were not cost-effective.
Many evaluations used cost-effectiveness designs.Although informative, these studies used a variety of disease-specific outcomes that are not directly comparable for interventions targeting the same problems or interventions across different diagnostic areas.Further, the use of clinical measures undermines the likelihood of detecting improvements that may be relevant to everyday life and general wellbeing, such as improvements in quality-oflife.This is particularly important in the case of parenting interventions that may have impacts on different areas of children's lives.Importantly, while there are established WTP threshold values for a QALY gained or a DALY averted, no such threshold value exists for disease-specific outcome measures, making it difficult to draw conclusions regarding the value-for-money of such interventions.To tackle these limitations, studies should include instruments that can capture health-related quality-of-life based on individuals' preferences.There are a few multi-attribute utility instruments (MAUI) available in the literature, which can be used in children [71], that make it possible to estimate QALYs and, thus facilitate value-for-money estimations.Currently available instruments are, however, limited to children older than seven years of age (unless proxies are used).A reason as to why multi-dimensional outcomes were not included in most studies in this review might be that children were of younger ages.The most common MAUIs may, however, not always be appropriate in some contexts, such as mental health, as they may not fully capture the elements of health-related quality-oflife most relevant to these children.Despite its usefulness and economic credentials, QALYs per se may also fail to capture important clinical improvements.Disease-specific instruments can be more relevant in such contexts.Future research should focus on employing and developing instruments that can capture meaningful changes for different populations.
The studies also adopted narrow costing perspectives, often including intervention and medical-related costs but lacking broader societal costs, such as educational sector costs, other societal services, informal care and productivity losses for parents.This narrow approach is likely to miss important impacts across different sectors of society.This is especially true for evaluations in child health, since many conditions have impacts across different sectors of society, and may also require the delivery of care in nonmedical settings, such as schools, home, and the community.For instance, antisocial behaviors are known to result in increased use of resources in different sectors of society, such as healthcare, educational and justice system services [12,72,73].Childhood anxiety disorders and child abuse also yield large costs to society, including indirect costs stemming from productivity losses of parents due to absence from paid work [74,75].Importantly, narrow costing perspectives limit the comparability with other interventions that may differently impact the use of resources, and may lead to inappropriate decision-making.It is, however, recognized that capturing the full scope of costs that may be impacted by a parenting intervention is a difficult task, given that many are likely to occur as the children get older.Additionally, other factors may pose difficulties to adopting a broader costing perspective, including the lack of routine data sources available for estimations of resource use beyond health care, as well as the financial burden of added data collection and the added burden of data collection on participants.With such difficulties in mind, the latest recommendations of the second panel on cost-effectiveness in health and medicine are that both a health care and if possible a broader societal presented are presented as reference cases [76].
Another important issue when evaluating parenting interventions are spillover effects, i.e., the impacts of the interventions not only on children themselves, but also on those who can be directly affected by changes in children's health and wellbeing-such as parents and siblings.Current guidelines from the U.S. [76], Canada [77], the UK [78], and the Netherlands [79] recommend the inclusion of family costs and "spillover effects" in economic evaluation when relevant.Including spillover effects in economic evaluation can change the value of an intervention [80].In a review of pediatric economic evaluations, the inclusion of spillover effects contributed to the cost-effectiveness of interventions being more favorable 75% of the time [81].In the current review, more than half of the studies included at least one type of family spillover costs (i.e., time costs or out of pocket costs) but only one included parent health outcome spillovers in the ICER estimate.
Existing economic evaluations of RCTs have quite limited time horizons, often below one year.Time horizon can strongly impact the results of an economic evaluation.On average, extending the time horizon of economic evaluations leads to more favorable estimates of value [82], which is important when the impacts of an intervention may extend into the future, as is the case for interventions in child health.Modeling studies can help address some of the issues of RCTs, through longer-term projections of estimated costs and outcomes, but should always use available evidence from real world data and assumptions.
Finally, it is important to stress the importance of planning for an economic evaluation upon study design to capture all important costs and outcomes, and use appropriate instruments to measure QALYs.This is not always the case, as a few evaluations included in this review appear to have been conducted on an ad hoc basis and lack inclusion of appropriate instruments to measure health outcomes and resource use.

Conclusions
The existing evidence suggest that parenting interventions are likely to be a cost-effective use of societal resources, with respect to preventing child externalizing and internalizing behaviors, as well as home-visiting programs to prevent child abuse and neglect.Family-based community programs targeting the treatment of obesity were not cost-effective.Future studies should aim to capture the full health and economic impacts of child health interventions.Investment in parenting interventions is value-for-money and worth serious consideration by decision-makers.

Fig. 1
Fig. 1 PRISMA flow diagram of study selection process

Table 1
Characteristics of the studies included Child and Adolescent Mental Health Services; TAU, treatment as usual; WC, waitlist control, y.o, year olds

Table 2
Summary of methods and results of the studies includedHealthcare provider with caregiver costs.Caregiver costs included the costs of foregone paid work time and out of pocket costs for transportation and child care c All costs converted to 2020 US$ from original currency using a conversion rate based on Purchasing Power Parities (PPP) for gross domestic product from http:// eppi.ioe.ac.uk/ costc onver sion/ defau lt.aspx b