Parenting practices and behaviours play an important role in shaping various aspects of children’s development (Ermisch, 2008; Marmot et al., 2010). Research has found that parenting practices characterised by positive displays of affection and consistent disciplinary strategies are related to a child’s psychosocial development, while parenting which lacks control or warmth is associated with a greater number of behavioural issues (Baumrind, 1978; Rose, Roman, Mwaba and Ismail, 2018). Parenting interventions can be an effective tool for improving parenting practices (Barlow, Parsons and Stewart-Brown, 2002) and are typically provided on an individual, case-specific basis. Yet offering parenting programmes on a universal scale has the potential to eliminate stigma surrounding parenting interventions, as well as serving as a preventative mechanism (Prinz, Sanders, Shapiro, Whitaker and Lutzker, 2016). Triple P is an example of a parent training intervention which adopts a population approach and uses a multi-method delivery system (Sanders, 2008). A systematic review of 101 Triple P studies found evidence of significant short and long-term effects for child well-being outcomes, as well as parenting practices and satisfaction, and parental relationships (Sanders, Kirby, Tellegen and Day, 2014).
The aim of this study was to evaluate the effectiveness of implementing Triple P at a population level in Ireland on the prevalence of children’s social, emotional, and behavioural problems as measured by the Strengths and Difficulties Questionnaire (SDQ; Goodman, 1997). The Triple P intervention was offered to all parents of children aged between four and eight years in two intervention regions over a thirty-month period between 2010 and 2013. The programme was delivered by the Longford/Westmeath Parenting Partnership (LWPP) using a partnership model comprising statutory, community, and voluntary sector organisations. Four of the five Triple P levels were provided including a media campaign, one-off seminars and discussion groups, and multi-week group sessions. Parents could choose how many and which levels of Triple P to participate in. A propensity score matching (PSM) differences-in-differences (DID) method was used to compare two intervention regions (Longford and Westmeath) and two comparison regions (Tipperary North and Tipperary South) matched on socio-demographic characteristics. By combining DID with PSM, we utilised the advantages of both techniques, as DID helps to differentiate out permanent (time invariant) confounders and PSM helps to capture transitory shocks (and thus balance the covariates). This method was first introduced by Heckman, Ichimura and Todd (1997). We first estimated the propensity score using a probit function to calculate the probability of being in the intervention sample based on observable covariates. The propensity score allowed us to match treated and untreated respondents according to one single metric. We used kernel matching which allowed a large amount of data to be retained, while bias was reduced by assigning larger weights to the untreated units which were the closest matches to the treated units. The Epanechnikov kernel was employed and the optimal bandwidth was determined using Silverman’s rule-of-thumb. Separate propensity score models were estimated for both time points. Next, we estimated the DID models using OLS regressions for the continuous outcomes and Linear Probability Models for the binary cut-off scores, and applied the average weights obtained from the propensity score in each case. Each DID model included a treatment status indicator (intervention or comparison sample), a time indicator (pre- or post-intervention sample), an interaction between time and treatment status, and a set of control variables. Please see the full paper for a detailed explanation of this method (Doyle, Hegarty and Owens, 2018).
A population survey was administered in both regions before (between February and April 2010) and after (between March and May 2013) programme implementation. Participants were sampled from 302 Electoral Districts across the four regions and a random probability sampling method was then used to select the sample of eligible households using quotas based on socio-economic group (Fives, Pursell, Heary, Nic Gabhainn and Canavan, 2014). The pre-intervention sample included 1,501 and 1,495 parents of children aged four to eight years in the intervention and comparison regions respectively. The post-intervention sample included 1,521 and 1,544 parents respectively. These samples represented seventeen to twenty per cent of all families with children between the ages of four and eight years in the counties.
Table 1 presents the mean scores of the intervention and comparison samples at the pre and post intervention periods, as well as the results of the PSM DID models. There were significant reductions in the prevalence rates of social, emotional, and behavioural problems in the two intervention regions compared to the two comparison regions. Children in the intervention sample experienced lower total difficulties, emotional symptoms, and conduct problems, and they were less at risk of scoring within the borderline/abnormal range for total difficulties, conduct problems, and hyperactivity. The programme reduced the proportion of range by 4.7 per cent for total difficulties, 4.4 per cent for conduct problems, and 4.5 per cent for hyperactivity in the total population. There were no impacts on peer problems or on positive behaviours as measured by pro-social behaviour.
This study provides some evidence that a universal system of parenting support may result in reductions in the prevalence rates of child social, emotional, and behavioural problems at a population level. This was achieved using Triple P, a universal parenting programme which was implemented at multiple levels by a collaborative partnership. The programme reduced the proportion of children classified as having borderline/abnormal problems, as well as reducing the overall level of difficulties. The results reported here differ from the one other comparable population study of Triple P in Australia which also adopted a quasi-experimental methodology and measured similar outcomes, and found that the effects were limited to total difficulties and emotional difficulties (Sanders et al., 2008). Possible explanations for differences in the number of treatment effects include programme developments over time. These results add to the evidence base for the science of population-based approaches to parenting.
A full-length version of this paper has been published as: Doyle, O., Hegarty, M. Owens, C. (2018) Population-based Parenting Programme to Reduce the Prevalence of Child Socio-emotional and Behavioural Problems: Differences-in-Differences Study, Prevention Science, 19 (6), pp. 772-81. This research was supported by the Children’s Research Network Prevention and Early Intervention Research Initiative Research Grant Scheme 2017-18.
[Footnote for table] Columns (1), (2), (3), (4) show the unweighted raw scores of the intervention and comparison groups in the pre and post-intervention periods. Column (5) shows the difference-in-difference impact from the kernel PSM DID models which includes a treatment status indicator (intervention or comparison), a time indicator (pre- or post-intervention), an interaction between time and treatment status, a set of controls (child age, child gender, respondent’s relationship to the child, Irish-born or not, employment status, educational attainment, social class, rurality, housing tenure, and aggregate ED), and matching weights applied.