Introduction

Young children who experience sustained socioeconomic and psychosocial adversity (such as parent unemployment, poor parent mental health, family violence, and substance abuse) are at risk of wide-ranging health and developmental inequities [1, 2]. These persist throughout life, resulting in lower educational attainment, poorer health and lower income [3]. Inequities are costly for the individual and society, and considered a major public health issue [1]. Despite global policy interest [1], it is extremely challenging to implement programs that improve the health and developmental trajectories of children and families experiencing adversity. Indeed, in high-income countries with universal healthcare the ‘inverse care law’ persists, whereby families with the greatest need are least able to access sufficient services or those of highest quality [4]. Starfield’s work into equitable primary care delivery [5], and Marmot’s focus on proportionate universalism as a core aspect of system change [6], suggest that health systems for children should allocate resources equitably and according to risk, with appropriate intensity and flexibility to counter the social determinants of health.

Nurse home visiting (NHV) is one of the most widely implemented early interventions for addressing maternal and child health inequities. NHV provides targeted and intensive delivery of child and family health care, with the potential to overcome barriers to service access for families experiencing adversity [7]. Randomised controlled trials (RCTs) of NHV programs in high-income countries with universal healthcare include the Family Nurse Partnership (United Kingdom) [8], Pro Kind (Germany) [9] and VoorZoorg (Netherlands) [10]. These programs have predominately focussed on early outcomes (up to child age 3 years) and demonstrated wide-ranging benefits across parenting, the home environment, and child language and behaviour [7]. Notably, the diversity of program content and measurement, and relative scarcity of RCTs, have impeded meaningful meta-analysis [7]. Overall, the reported benefits of NHV tend to be variable and modest [7] which, at the population level, make sense when considering the multi-dimensional nature of the programs, the variation in nurse practice, and the complex experience of adversity. In this context, the benefits of NHV have been noted as one of the few public health programs that can make a measurable difference for families and children in the earliest years of life [11].

Few NHV programs have been rigorously evaluated beyond children’s third birthdays [7]. The most recent systematic review of NHV RCTs was conducted by Molloy et al. (2021) [7], and focused on outcomes measured from pregnancy to child age 5 years. Of 30 articles describing the results of seven NHV programs, only two reported evaluations to 4–5 years. Of these, the Nurse Family Partnership program improved children’s language at 4 years [12], and the Minding the Baby program reduced children’s externalising behaviours at 5 years [13]. In a separate RCT of the NFP program, Olds et al (2004) also reported improved child behaviour outcomes at 6 years [14]. However, these studies were conducted in the United States, where there is no universal health insurance and variability in the types of healthcare provision to serve as a comparator. As such, the findings are more limited in their application when considering the longer-term efficacy of NHV embedded in universal healthcare. In the context of high-quality child and family health (CFH) services, the comparative benefit of NHV may be different or less apparent, particularly over the longer-term. Only the German Pro Kind program has been evaluated for longer-term outcomes within a universal healthcare context [15]. At child age 7 years, Pro Kind demonstrated improvements to child behaviour and maternal mental health outcomes and reduced abusive parenting, with effect sizes (ES) ranging 0·19 to 0·25.

In Australia, a country with universal healthcare provision, “right@home” is the largest and only multi-site RCT of NHV [16]. Conducted in partnership with two Australian state governments, the right@home NHV program was offered to women from pregnancy (recruited in 2013–14) to their subsequent child’s second birthday (2015–16). At the program’s end, the trial demonstrated intervention benefits across the three primary outcome domains of parent care, responsivity and the home learning environment [17]. Evidence of benefits for maternal mental health and wellbeing also emerged at the 3-year-old follow-up, one year after the program ended [18]. Given the limited longer-term data on NHV effectiveness for countries with universal healthcare, we extended the right@home trial follow-up to child ages 4 and 5 years (2017–19) with the intention to directly assess children’s language and learning outcomes up to the year before school-entry. Following the program logic underpinning the right@home program [19], we anticipated that the parenting-related benefits evident at 2 years, and the maternal mental health benefits evident at 3 years, would translate to emerging benefits in children’s learning and development as well as sustained benefits for parenting and maternal mental health. We hypothesised that, at child ages 4 and 5 years, compared with families who received the universal CFH nursing service (usual care), families who received the NHV program (the intervention) would demonstrate improvements in children’s language and learning, general and mental health, maternal mental health and wellbeing, parenting and family relationships.

Methods

Outcome measures (Table 1)

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Description of outcome measures assessed at 4 and 5 years.

https://doi.org/10.1371/journal.pone.0277773.t001

Outcomes were identified a priori, across the domains of children’s language and learning, general and mental health, parenting and family relationships, and maternal mental health and wellbeing. Given the complex nature of the right@home program, which was designed to improve multiple outcomes across child and parent domains, prioritising a single outcome (as is common in clinical RCTs) is likely to understate the effect of the intervention [23]. We therefore examined multiple outcomes, including direct assessment of child language and learning outcomes (receptive and expressive language, phonological awareness, attention and executive function). By examining the direction, magnitude, and confidence intervals of estimated effects for multiple outcomes, we aimed to comprehensively evaluate the evidence for treatment benefits [24].

Results

Fig 1 shows that of the 722 women originally enrolled in the right@home trial, 465 (64·4%) provided data at 4 years (collected from 25 July 2017 to 3 January 2019) and 426 (59·0%) at 5 years (collected from 3 July 2018 to 14 January 2020). Table 2 shows that compared with women lost to follow-up at 5 years, those who participated in the intervention experienced less adversity at baseline (pregnancy) in terms of mental health, employment, ever having a drug problem or reporting family violence in the preceding year. Among those retained, baseline characteristics were similar between the intervention and usual care arms, although more children were female in the retained intervention families compared to usual care. S1 Table shows that the distribution of participant characteristics by follow-up status and trial arm was similar at 4 years.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Consort diagram.

https://doi.org/10.1371/journal.pone.0277773.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Baseline characteristics according to follow-up status (retained or lost to follow-up) at child age 5 years ^a.

https://doi.org/10.1371/journal.pone.0277773.t002

The adjusted regression analyses (with MI/IPW) show the direction of effects across child (Fig 2 and Table 3) and maternal (Fig 3 and Table 4) outcomes. The Figures present the direction, magnitude, and confidence of each estimate, allowing the overall evidence of treatment effect to be interpreted. Considered collectively, the estimated group differences for child and maternal outcomes favoured the intervention arm. The statistical evidence was stronger for parents than for children, and stronger at 5 years than 4 years. Effect sizes in favour of the intervention ranged from 0·01 to 0·27.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2.

Results of adjusted regression analyses comparing the two trial arms on continuous (top box) and categorical (bottom box) child outcomes at 4 and 5 years, using multiply imputed and inverse probability weighted data.

https://doi.org/10.1371/journal.pone.0277773.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3.

Results of adjusted regression analyses comparing the two trial arms on continuous (top box) and categorical (bottom box) maternal outcomes at 4 and 5 years, using multiply imputed and inverse probability weighted data.

https://doi.org/10.1371/journal.pone.0277773.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Results of adjusted regression analyses comparing the two trial arms on child outcomes at age 4 and 5 years, using multiply imputed and inverse probability weighted data.

https://doi.org/10.1371/journal.pone.0277773.t003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Results of adjusted regression analyses comparing the two trial arms on maternal outcomes at child ages 4 and 5 years, using multiply imputed and inverse probability weighted data.

https://doi.org/10.1371/journal.pone.0277773.t004

For children, there was statistical evidence of improved mental health outcomes at 5 years (reduced SDQ Externalising [ES: 0·14; 95% CI: -0·00 to 0·29] and reduced SDQ Total Score [ES: 0·20; 95% CI: 0·06 to 0·34]). Although the estimates favoured the intervention group, there was no statistical evidence of intervention effects for the directly assessed language and learning outcomes (receptive and expressive language, phonological awareness, attention, and executive function) at either time point.

There was evidence of intervention benefits across multiple outcomes of parenting at one or both time points (warmth at 5 years [ES: 0·18, 95% CI: 0·02 to 0·33], reduced hostility at 4 years [ES: 0·14, 95% CI: -0·01 to 0·28], efficacy at 5 years [ES: 0·21, 95% CI: 0·05 to 0·37], reduced parent-child conflict at 4 years [ES: 0·15, 95% CI: 0·01 to 0·29], regular mealtimes at 5 years [OR: 3·79; 95% CI: 1·81 to 7·92], regular bedtimes at 5 years [OR: 2·42; 95% CI: 1·43 to 4·10], and bedtime routine at 5 years [OR: 1·56; 95% CI: 1·08 to 2·26]). There was also evidence of intervention benefits to maternal mental health and wellbeing outcomes (reduced DASS stress symptoms at 5 years [ES: 0·12; 95% CI: -0·08 to 0·31, which was also reflected in the dichotomous stress variable], personal wellbeing at 4 and 5 years [ES: 0·17; 95% CI: 0·08 to 0·26, and ES: 0·27, 95% CI: 0·06 to 0·48], quality of life at 4 and 5 years [ES: 0·14, 95% CI: -0·02 to 0·30, and ES: 0·15, 95% CI: -0·01 to 0·32, which was also reflected in the dichotomous global health variable], and self-efficacy at 5 years [OR: 1·61, 95% CI: 1·13 to 2·31]). Intervention group mothers were more likely to report absence of emotional abuse from a partner at 5 years [OR: 1·72; 95% CI: 0·96 to 3·06]. The only outcome that showed evidence in favour of the usual care arm was that fewer intervention group mothers reported currently studying at 5 years [OR: 0·60; 95% CI: 0·40 to 0·92].

Similar associations were also identified in the complete cases analyses for child (S2 Table) and maternal (S3 Table) outcomes. In addition, there was evidence of benefit across the directly assessed child outcomes of receptive and expressive language (CELF Sentence Structure [ES: 0·21], Word Structure [ES: 0·15] and Core Language [ES: 0·17]), phonological awareness (SEAPART Total Score [ES: 0·21] and parent reported child mental health outcomes (reduced SDQ Internalising [ES: 0·20] and SSIS Self Control [ES: 0·20]) at 5 years in the complete cases analyses. Estimates were of similar magnitude to those in the MI/IPW analyses supporting a lack of bias introduced by attrition and missing data. Confidence intervals were generally narrower for complete cases reflecting the additional uncertainty introduced by using missing data techniques such as multiple imputation.

Discussion

This study reports the only longer-term evaluation of an Australian NHV trial designed to support women experiencing adversity from pregnancy to their children’s second birthday and embedded in the universal CFH service. When children were 4 and 5 years old, which was two and three years after the intervention ended, there was evidence of modest benefits of the NHV intervention across multiple domains including children’s mental health, parenting and family relationships, and maternal mental health and wellbeing. While there was no evidence of benefits in children’s language or learning, there was a pattern across estimated group differences of improvements (albeit small) in favour of the intervention.

When considered with the earlier waves of follow-up, this study found emerging benefits for child behaviour (which were not evident when first assessed at 3 years) [51], and sustained benefits across parenting, family functioning and maternal mental health outcomes which increased at 5 years [18, 51]. At 5 years, there was also evidence of reduced intimate partner emotional abuse. This is notable since the intervention was a parenting intervention, rather than one that specifically targeted mental health or intimate partner abuse. Given the dearth of effective preventative interventions targeting inequitable outcomes, our research strengthens the evidence for NHV as an important part of a proportionate universal healthcare system.

The benefits identified at 4 and 5 years are consistent with the intervention focus and curriculum delivered by nurses [22]. As described previously [22], maternal mental and physical health and wellbeing support were the most frequently delivered program components, provided in over 85% of nurse visits. Discussions about relationships, including partnership issues and social support were also common, recorded for 50–70% of visits, with the topic of domestic violence recorded in a quarter of all visits [22]. Other areas of focus included care planning, caregiver aims and goals, infant feeding, growth and development, and housing and environment (discussed in at least half of the nurse home visits). The fidelity data align with the trial’s program logic [19], which articulated that early parental benefits would generate child developmental gains, which appear to be emerging at 5 years. These findings strengthen the international NHV evidence base by linking the direct inputs of the intervention with longer-term outcomes. For example, the emerging reduction in intimate partner emotional abuse (noting the substantial intervention input on relationships) is important given the lack of effective preventive interventions in this space [52].

Our findings also align with the small number of existing studies that have examined the longer-term effects of NHV programs. The recent 7-year-old follow-up of the Pro Kind program evaluated in Germany’s universal healthcare setting also reported benefits for children’s behaviour, reductions in adverse parenting and improved maternal mental health outcomes, but no evidence of benefits for children’s school performance outcomes [15]. In our study, effect sizes for estimated group differences with statistical evidence ranged 0·12 to 0·27, and for Pro Kind ranged 0·19 to 0·25. While the US Nurse Family Partnership and Minding the Baby programs were evaluated in a different service context, they too reported benefits to child behaviour at 4 to 6 years with available effect sizes ranging from 0·06 to 0·17 [13, 14]. NFP additionally reported no impact on children’s executive functioning at 4 years and mixed evidence for their academic outcomes (again with equivalent effect sizes) at 6 years [12, 14]. These similar outcomes were achieved with distinct programmatic inputs that may be important for policy scaling, such as differing foci in the curriculum and number of visits (e.g. right@home delivered an average of 23·2 home visits, compared with NFP which delivered an average of 7 antenatal and 26 postnatal visits, and Pro Kind which delivered an average of 32.7 home visits).

The strengths of this study include the rigorous, real-world, and prospective RCT design, and the comprehensive, longitudinal follow-up. Although loss-to-follow-up increased over time, we retained a relatively high proportion of participants (59% over a 6-year study duration) given the high levels of adversity and re-enrolment processes. In comparison, by their respective 2-year follow-up, the UK Building Blocks study retained 71% of their cohort for self-reported outcomes [8], and Pro Kind retained less than 50% [9]. Recognising that NHV trials are susceptible to participant attrition, which could introduce non-comparability between the trial arms, we applied the best practice methods of using a combined MI/IPW approach to address missing data [50]. Consistency between the complete case analysis and analysis with missing data accounted for, indicates that the intervention benefits evident in both sets of analyses are unlikely to be artefacts.

The balance of evidence across the outcomes (as illustrated in the forest plots) indicate beneficial outcomes overall. However, a limitation of the MI/IPW methods is that they introduce uncertainty into the estimates. As such, all estimates had wide confidence intervals and few outcomes reached statistical significance, meaning it is difficult to closely identify the specific effects on each given outcome. Missing data for all outcomes were also treated using the same approach, where some outcomes may have different reasons for missingness. For example, questions about whether mothers were experiencing intimate partner emotional abuse could only be administered when there were no other family members present, resulting in higher proportions of missing data than other measures. It is also important to consider why group differences were evident for outcomes reported by mothers but not for those that were directly assessed. Ongoing follow-up of the right@home cohort when children start school will enable comparison of teacher and parent reported data, and reveal whether developmental impacts emerge at a later age.

In thinking about the future of the NHV evidence base, we note that our limited sample size and power for detecting small group effects will continue to be an ongoing challenge, as it is for other studies. A substantially larger-scale evaluation would be needed to identify longer-term benefits more definitively. Lifetime cost-benefit modelling of NFP in the US shows that benefits accrue to participants and taxpayers over a child’s lifetime, producing positive returns on investment [51, 53], which is similarly beyond the scope of almost all research trials. While cost evaluations in universal healthcare settings are limited, our economic evaluation of right@home at child age 3 years demonstrated consistent findings; NHV involves high up-front costs that will be recouped if, and as, benefits emerge over time including those from early identification and intervention [51]. Given the demonstrated benefits of NHV, policy evaluations using designs such as stepped-wedge trials alongside program implementation should be prioritised into the future with a strong focus on implementation rather than smaller RCTs [54].

This study has demonstrated that it is possible to deliver a proportionate and targeted program within a universal CFH service that can provide important protective benefits to children and families experiencing adversity. Although modest effects delivered at scale can have a significant population impact, even the best outcomes of NHV highlight that a single intervention cannot fully redress inequities in health and development experienced by families. A sustainable and effective response is likely to require sustained intervention across the life course, identifying effective interventions and their ideal time point(s), duration, and intensity to maximise impact [55].

Conclusion

Our findings suggest that the right@home NHV program offered sustained benefits to parenting, family functioning and maternal mental health and wellbeing for families experiencing adversity, when embedded into Australia’s universal CFH service. Given the dearth of well tested preventive interventions, right@home offers a potentially effective option for a proportionate universal system that hopes to deliver support and intervention relative to need. Now, more than ever, as inequities likely increase in response to the COVID-19 pandemic, programs like right@home are needed to support families who could benefit most.

Supporting information