Dose response relationship between program attendance and children’s outcomes in a community based weight management program for children and their families

Childhood obesity can lead to lifelong overweight and obesity with adverse health effects throughout life. A recent analysis [1] of worldwide data has shown that childhood obesity rates have plateaued in developed countries, albeit at high levels, and are increasing in developing countries. A concerted approach is required to tackle childhood obesity that includes community based programs to manage childhood obesity, interventions in schools and other settings and media campaigns to promote healthier lifestyles and policy changes to influence health behaviours [2]. Community based interventions at the family level are one of the most effective in managing childhood obesity [3] and a systematic review found them to be more effective than school based interventions for children under the age of 12 years [4].

The Go4Fun program in New South Wales (NSW), Australia, described elsewhere [5, 6], is a family based community weight management program for overweight and obese children aged 7–13 years that is delivered once per week [7]. It is a mandatory requirement of the program that a parent or a carer attend the program sessions with their children. The program is a community-based translation of the MEND (Mind Exercise Nutrition Do it) program in the UK which is delivered in association with clinical services [8]. It has since been further modified to adapt to Australian settings and in response to ongoing quality assurance and research findings [5, 7]. The most significant modification was changing program delivery from twice per week to once per week to reduce participants burden and service costs. A cluster randomised controlled trial showed that the condensed once per week version was as effective as twice per week version of the program in achieving health and behavioural outcomes [7]. No program changes were made to the once per week version of the program for the purpose of the current study.

Understanding the level of attendance that is likely to yield optimal participant outcomes would help program staff develop strategies to encourage participants to attend at least the minimum number of sessions that is likely to deliver beneficial lifestyle and health outcomes. The importance of assessing the dose-response relationship between attendance and outcomes for programs targeting childhood obesity has been highlighted previously [9], but it has also been acknowledged that dose-response relationships are difficult to discern in controlled studies because of limited sample size and variability in data [10].

In this paper, using routinely collected pre and post program data for the Go4Fun program between July 2013 and December 2016, we describe the impact of the number and type of sessions attended on BMI zscore and nutritional behaviours at program completion to determine the number and type of sessions likely to generate optimal program outcomes.

Secondary analysis was conducted on routinely collected de-identified Go4Fun program data, which includes anthropometric measurements of the participating child and the attending parent and parent reported nutritional, physical activity and sedentary behaviours. Go4Fun program is delivered over 10 weeks during school terms (January to December, with four terms). Groups typically comprise of 8–12 children. The program is delivered by the local health districts (LHDs), geographical administrative regions for health services in NSW.

All program participants for whom pre and post data were available between July 2013 and December 2016 were used for this study, which includes data for 265 children from a pragmatic incentives study [11] to assess the effectiveness of incentives to families in improving program participation and outcomes (results not yet published). No changes were made to the program structure and delivery approaches for the incentives study.

A total of 6366 eligible children registered for the program within the study period of which pre-program data were available for 5389 participants (84.7% of registrants) and post-program data for 3090 participants (57.3% of those with pre-program data). Program delivery occurred at 426 sites, with the number of participants ranging from four to 22. The intracluster coefficient (ICC) for BMI z-scores was found to be 0.13.

The average age of the children who registered for the program was 9.8 (±1.9) years, 50.6% of whom were females (see Additional file 1: Table S1). Most participants were brought to the program by their mothers (n = 4327, 72.6%). Children who did not attend the post-program measurement session attended 4.2 (±2.3) sessions on average whereas the children for whom the post program data were available attended 8.8 (±1.4) sessions on average. The children who completed the Go4Fun program were more likely to be from families with post-school qualified mothers and those that lived with more than one parent or carer. There were no differences in the other demographic characteristics and the health and behaviourial measures between the children who attended or did not attend the post program measurements (Additional file 1: Table S1).

Children with post program data who attended fewer than five sessions (n = 38) were excluded from further analysis as the low numbers did not allow a reliable assessment of the association between attendance and outcomes. For participants who attended at least five sessions, there was a statistically significant change in BMI z-score and other outcome measures at post program compared with pre (p < 0.01 for all outcome measures) (Table 2). The number of sessions attended showed weak correlations with BMI z-score and fruit and vegetable intake and no correlations with time spent on physical and sedentary activity (Table 2).

The multivariate analysis of variance (F = 3.68, p = 0.03) and the subsequent multivariate regression showed that the only variable likely to influence the BMI zscore was highest qualification of mother. (Table 3). Children of mothers without university degree qualifications were likely to achieve significantly lower levels improvements in BMI zscore (β =0.02, p < 0.01) at program completion compared to the children of mothers who were university degree qualified. No effects were seen for number and type of attended sessions, age and gender of child, parent BMI at pre-program and household type.

Our study, using routinely collected data from large scale routine implementation of the Go4Fun program, supports previous reports that participants of the scaled up childhood obesity programs based on the MEND program achieve positive health and behaviourial outcomes at program completion [22‐25]. Our study adds to the evidence that large-scale weight management programs for families of overweight and obese children are a useful tool in tackling obesity in children, in line with findings from previous reports of MEND based and other interventions [3, 22‐26].

An important finding in our study was the association between maternal educational level and program outcomes. Participants that had mothers without university qualifications benefited incrementally from attending more sessions until seven sessions after which the impact of the program plateaued, whilst other participants seem to achieve better BMI z-score outcomes by attending fewer sessions. It has been reported in other areas of health care that families with low level of parental literacy are less likely to fully participate in or adhere to programs like Go4Fun [27, 28] and achieve worse health care [29] and program outcomes [30] which subsequently lead to health inequities [31]. Specifically for the Go4Fun program, strategies need to be developed to encourage families with low levels of parental literacy to attend at least seven sessions. Families who speak a language other than English at home may also benefit, but our analyses did not find language spoken at home was associated with attendance. This target should be achievable as more than half of the families with mothers without university degree qualifications in Go4Fun already attend at least seven sessions.

There was no distinctive pattern in our study of participants attending either a session which was based on learning information or obtaining knowledge e.g. a session learning about fats and sugars or a session around application of knowledge e.g. supermarket tour. Session type did not influence the program attendance or outcomes for Go4Fun families. This may be because attendance is often determined by factors outside of the program schedule (for example, illness, or competing time demands) and material covered in the sessions can be caught up or learned from the program resource material. Interest in program content is one of the main predictors of attrition in paediatric weight management programs [32, 33] along with other competing priorities and logistical factors [33]. From a program implementation perspective, this means that it is crucial to ensure participants remain interested in the overall program content area but it may not be necessary to prioritise some sessions over others.

In a broader context, our finding highlights the strengths of routinely collected robust program data in monitoring and evaluating weight management programs delivered at scale to identify program implementation strategies that can be key influencers of program outcomes [34]. Conversely this is not usually possible with experimental data due to insufficient sample size. Nonetheless, the use of routine program data for assessing outcomes and using clustered data have limitations. Although recent studies have shown that measurements by trained staff are sufficient for research purposes and are reliable [35, 36], we could not validate the accuracy of the collected anthropometric data. Further, self-reported/parent-reported physical activity, sedentary behaviour and dietary intake has some limitations, as these behaviours can be difficult to recall. However, child recall of these behaviours is very poor, and at least the parents are able to report consistently (at the beginning and end of the program) about what they were able to observe. The ICC in our study was considerably higher than the values reported for other prospective clinical and population based studies with cluster designs. The high ICC for BMI zscore in our sample is perhaps not surprising given the mandatory program requirement that all participants need to be overweight or obese. Nevertheless, the high ICC reinforces the need for large sample sizes when assessing program outcomes.

Given the large sample size however, it can be expected that the effect of any measurement errors in the final outcomes is likely to be small. Also, it is not possible from our study to establish cause-effect associations for program outcomes, and we could not ascertain the reasons for families withdrawing from the program. Despite these limitations, we have been able to gain insights into how the program benefits can be optimised for participating children.

Our study has been able to show that children from families with mothers without university degree qualifications need to attend more program sessions than those with university degree qualified mothers to achieve similar level of health and behaviourial outcomes in family based weight management programs.