Costs and consequences of the Family Nurse Partnership (FNP) programme in England: evidence from the Building Blocks trial

Sample

The Building Blocks trial was a two-arm pragmatic, non-blinded, parallel-group, randomised controlled trial (RCT) which recruited within a community midwifery setting at 18 partnerships between LAs and primary and secondary care organisations in England. A total of 1645 participants were randomised in the study. The present analysis is based on the 1618 participants recruited to the trial that did not later become ineligible (e.g. due to miscarriage) or withdraw their consent. Participants in the Intervention arm (n=808) received home visits from the FNP nurse during their pregnancy and in the two years following childbirth. Participants allocated to the usual care arm (n=810) received care from the local maternity and health visiting services in line with usual practice. Details on recruitment and follow-up rates are described in full elsewhere¹. Informed consent was obtained from all individual participants included in the study.

Outcome measures

The primary outcome measures for the trial were tobacco use at late pregnancy (34–36 weeks’ gestation), birth weight, emergency attendances and hospital admissions for the infant within 24 months of birth, and the proportion of women with a second pregnancy within 24 months post-partum. There were also a large number of secondary outcomes across a number of domains; socio-economic, maternal health and well-being, health behaviours, pregnancy and birth, social support and the use of services. For all primary outcomes a 97.5% confidence interval (CI) and p-value was presented; for all secondary outcomes a 95% CI and p-value was presented. The outcome measures are described in full elsewhere¹.

Economic data collection

Data for outcomes and resource use were collected by self-reported questionnaires at various time-points throughout the trial, specifically; baseline, late pregnancy (34–36 weeks gestation) and 6, 12, 18, and 24 months postpartum. Baseline and 24 month data were collected by face-to-face interview by a locally based researcher whilst data at other time points were collected via telephone by qualified telephone interviewers. Additionally, data related to the use of health care services for each trial participant were collected from Hospital Episode Statistics (HES) via NHS Digital and primary care (general practitioner (GP) records). Birth data were collected by records abstraction from maternity records.

Resource use

The mean numbers of the health and non-health care items per woman are presented for both groups for the duration of the trial in their respective units, e.g. mean number of GP visits, average number of weeks in education etc. The difference between the two groups for these items is also presented. The totals for each item (e.g. total number of GP visits) have been calculated for each participant in both groups over the duration of the trial (i.e. from baseline up to the 2-year interview).

Resource use is averaged across all women in each arm leading to small mean numbers for resource use. Median, minimum and maximum values are reported to highlight the highly skewed nature of resource use. Costs were largely associated with only a minority of women using each resource. Resource use and associated costs are reported in terms of the mother only. All consequences of the intervention, for both mother and child, are reported.

Unit costs

Unit costs applied to the resource use were retrieved from several sources. Unit costs pertaining to health care resource use (Table 1) were sourced from NHS reference costs⁶ and Personal Social Service Research Unit (PSSRU) Unit Costs of Health and Social Care 2013⁷. However, as no unified document currently exists to provide unit costs for the non-health related items, these were sourced from other relevant sources such as governmental websites, research documents and reports. Given that costs were taken from a range of sources the years of pricing vary for each item. These are reported in Table 2. No attempt was made to discount/ inflate these items to a consistent rate as appropriate inflation rates for this type of resource are not available. It is thus acknowledged that current figures per resource use may differ slightly from those documented.

Missing data

For the purpose of the cost-consequence analysis, multiple imputation analysis was considered unnecessary as the model serves only to highlight likely resource use and the average costs per resource. This decision was further vindicated by the small number of women who reported using each of the resources, which in some cases were less than five women. Such small numbers of affirmative responses would prevent imputations from running effectively and result in high error rates. Instead, several assumptions were made to reduce the amount of missing data. Where a particular answer to a question meant that sub-questions were skipped, a value of zero was attributed to the sub-questions. For example, respondents were first asked a yes or no question as to whether they had utilised a particular resource, then how often following an affirmative response. Where respondents stated that they had not used a resource, frequency of use was assumed to be zero. Conversely, where a positive response was noted but not frequency value was given, a value of one was assumed as a conservative estimate. Due to the high number of variables and low number of complete cases, it was not considered pragmatic to summarise each variable by complete case and adjusted values.

Statistical analysis

The mean resource use was calculated for each resource use item according to the appropriate unit (e.g. mean number of days). The incremental difference between trial arms was calculated for each item of resource use by subtracting the values of participants allocated to receive usual care from the values of participants who were allocated to receive the intervention (FNP). Mean unit costs were calculated by item for each of the trial arms by multiplying the resource use by an appropriate unit costs (see Table 3 for unit costs). Resource use costs were summed by type, e.g. health related resource use or non-health related resource use. All analyses were conducted in Stata version 12⁸. The consequences presented in Table 3 are the primary outcomes of the Building Blocks trial and full details of how these were calculated are presented in primary trial paper⁹.

Costs

Little difference was observed in terms of average health care resource use across the majority of services though consistently small differences were found between trial arms in favour of FNP. The largest differences in terms of resource use were the average number of A&E attendances and the mean number of health visitor visits received, both of which, as anticipated, were lower for women receiving the FNP intervention compared to women under usual care.

As only small differences in average resource use were observed, the average per woman spend on each resource reflected only slightly notable differences. Overall, average total health-related resource use costs were slightly lower for the FNP arm compared to usual care. Lower average costs were found in the FNP arm in relation to inpatient care, outpatient attendances not related to maternity services, A&E attendances, and midwife and health visitor visitations. The largest difference in costs was found in relation to inpatient care requiring one or more nights in hospital. In this instance, the mean difference between arms was £307 with lower costs being associated with FNP recipients. A difference was also found between the average cost per participant in terms of community-based maternity and early years services, specifically health visitor and midwife visitation. On average £82 less was spent per woman receiving the FNP intervention compared to usual care for health visitor visits (both home and clinic based) whilst £21 less was spent per woman on midwife visits (both home and clinic based) reflecting a mean reduction in spend of just over £100 per woman on these key early years services. The number of health visitor visits received by women was lower in the FNP arm than the usual care arm as visits from FNP nurses were received in place of standard health visitor visits. Consequently, though the cost attributed to health visitor visits appears reduced, there is the additional cost of the FNP nurses. The cost of the health visitors provided in usual care would thus not be saved through the FNP intervention but would be spent instead on specific FNP trained nurses. Only minor differences were noted in terms of cost associated with inpatient day admittances not requiring an over-night stay (-£6.51), non-maternity based outpatient attendances (-£4.86), and A&E attendances (-£5.73), with all being in favour of FNP. Omitting the intervention costs, a total of £8, 872 an average was spent on health care for women receiving the FNP intervention compared to a total average of £9, 162 per woman in the usual care arm, a difference of £290 per woman between the two arms.

In addition to health care resource use, there was very little difference in the non-health care resource use between the two groups. The average cost of non-health related resource use for women receiving the FNP intervention was £1,866 compared to £1,738 for women under usual care, an average difference of only £128 per woman in favour of the control group. In general, reported uptake of all non-health related resources was fairly low in both groups as can be seen through the mean and median values presented in Table 3. For the majority of resource items, the median resource use for both groups is zero, confirming a low uptake. Mean values are thus influenced by only a minority of women.

Participation in education and training was also similar between groups for both mainstream schooling and dedicated support units with little difference in resource use or mean costs.

In terms of supportive services, children’s centres, such as Sure Start centres, were the most widely used service by both groups, though this averaged at only 2 visits per woman in each of the trial arms. Mean costs of children’s centre use was slightly higher in the FNP arm. This is likely a consequence of outlying values as the maximum number of visits was considerably higher in the FNP arm compared to usual care (38 and 29).

Resource use of childcare was found to be particularly low at all time-points and overall mean resource use per woman, within both groups, was particularly low. Slightly higher costs were found in FNP women in line with the marginally higher rates of resource use. The most widely used type of childcare was crèche/day nursery within a school or college though the mean number of weeks used throughout the follow-up period was very low at only 1.35 weeks for FNP women and 1.10 weeks for usual care women. As this resource use was low the total resource use was low resulting in a low average cost per woman of £5.79 per FNP woman and £4.71 per usual care woman though the cost for women actually using childcare would be much higher.

Resource use for foster care and temporary accommodation was also low in terms of the number of women reporting usage of these resources for both groups of women. Average costs are thus driven by a minority of women. Average costs for temporary accommodation were higher for FNP women than usual care women for teenage parent accommodation, supported accommodation and mother and baby units due to increased use of these by FNP women. The largest average difference between groups was for teenage parent accommodation, with FNP women costing on average £75.73 more than usual care women; again this was due to higher levels of resource use in the FNP group.

Consequences

Primary outcomes. No difference was observed in the rate of smoking in late pregnancy between the two arms (adjusted OR: 0.90, 97.5% CI: 0.64 to 1.28). There was no difference in the reported number of cigarettes smoked during late pregnancy for participants classified at baseline as smokers (adjusted difference in means, FNP-Usual Care: 0.119 cigarettes, 97.5% CI: -0.73 to 0.97). On average, birth weights were marginally higher in the intervention arm compared to the control arm though this was not statistically significant (adjusted difference in means, FNP-Usual Care: 20.75 grams, 97.5% CI: -47.73 to 89.23). There was no difference in the proportion of women reporting a second pregnancy within the two years of their first child’s birth (adjusted OR 1.01, 97.5% CI: 0.77 to 1.33). Rates of child emergency attendances and admissions for any reason in secondary care prior to their second birthday were higher in the intervention arm though the difference was not statistically significant (adjusted OR 1.32, 97.5% CI: 0.99 to 1.76). There were no differential effects due to age, deprivation, participation in employment, education or training, or basic life skills for any of the primary comparisons.

Secondary outcomes. Some benefit of the intervention was observed in terms of maternally rated language developmental delay in the children at both 12 (OR 0.50, CI: 0.35 to 0.72) and 18 months (OR 0.66, CI: 0.48 to 0.90). This was further reflected in the Early Language Milestone assessment at the end of the trial period which showed higher scores on average for children in the intervention arm (adjusted difference of means, 4.49, CI: 0.52 to 8.45) indicating more advanced levels of language development. Benefits of the intervention were also observed for breastfeeding intentions (adjusted OR 1.32, CI: 1.02 to 1.70) though this did not translate into a difference in the actual proportions of mothers initiating breastfeeding between the two arms.

A greater proportion of children in the intervention arm attended an Emergency Department for an injury or ingestion at 6 months (adjusted OR 1.52, CI 0.86 to 2.70) and 12 months (adjusted OR 1.16, CI 0.92 to 1.46), were referred to social services by two years (adjusted OR 1.27, CI: 0.93 to 1.73) and had a safe guarding event recorded in their GP record (adjusted OR 1.85, CI: 1.02 to 2.85) compared to children in the control group. These differences were not statistically significant.

No evidence of a difference was observed for any other maternal or parenting and child outcomes assessed in the trial.

At two years, mothers in the intervention arm reported lower rates of not being in employment, education or training (NEET) than the control arm, though this was not statistically significant. A larger proportion of participants in the intervention arms reported a maximum level of social support at 18 months postpartum (25.7%) compared to those in the control arm (20.3%) with a similar difference being maintained at 24 months. A small difference between the arms was observed over the whole follow-up period (adjusted OR 1.5, CI: 1.06 to 2.12). Fewer participants in the intervention arm reported ever being homeless during the study period compared to the control arm (adjusted OR 0.76, CI: 0.55 to 1.05) and higher levels of self-efficacy were observed in the intervention arm (adjusted mean difference 0.44, CI: 0.10 to 0.78).

No evidence of a difference was observed for any other parental life course outcomes assessed in the trial.

Underlying data

The Building Blocks Trial dataset comprises data from multiple sources including third party. The Hospital Episode Statistics (HES) data, which forms a substantial component of the present analysis, is supplied under licence from NHS Digital. The HES dataset is not currently available for analysis to researchers outside Cardiff (the lead institution) due to the information provided to participants at consent (which identified which institutions were acting as data controller / processors).

Researcher can apply for access to the full dataset by contacting Cardiff Centre for Trials Research (CTR) (see: https://www.cardiff.ac.uk/centre-for-trials-research/about-us/data-requests). The restriction on HES data (originally supplied via NHS Digital) applies. CTR will consider whether section 251 approval would enable release of participant identifiers to apply directly to NHS Digital for the data used in the trial.

Extended data

The results of the Building Blocks trial are presented in full at: https://www.cardiff.ac.uk/__data/assets/pdf_file/0009/504729/Building-Blocks-Full-Study-Report.pdf

This publication includes the full CONSORT statement for the trial on page 564.