Benefits of Incentives for Breastfeeding and Smoking cessation in pregnancy (BIBS): a mixed-methods study to inform trial design

Engaging service-user groups

At the grant application stage, service-user engagement was initiated by approaching maternity services, primary care and children’s centre managers to select two thriving, but diverse, groups operating in areas of high deprivation in Aberdeen and Lancashire. Managers identified group representatives to negotiate involvement with the study: a health visitor – Wendy Ratcliffe (Aberdeen) – and a community centre worker – Helen Cook (Blackpool) – both named in the grant application. Areas of high deprivation were considered most likely to be able to provide insights from service users who had experience of the target behaviours for change: smoking in pregnancy and formula feeding. 53 Moreover, the selected groups were considered to be particularly well suited to collaboration on the BIBS project: the Aberdeen group is user led, is successful in generating some funding and provides a café and crèche as incentives to attend; the Blackpool group is located in a local authority-run children’s centre. The settings for these groups were summarised in Chapter 1 (see Box 1) and the group profiles are summarised in Table 1.

The group representatives were sent drafts of the grant proposal to discuss with the groups and they provided feedback during the planning stages. They negotiated reimbursement for their group’s refreshments and crèche costs as part of the grant application. Although the service-user collaborators were independent or local government representatives, rather than NHS groups, we considered it preferable to gain ethics committee approval before active engagement, particularly because the groups had not been involved in research before. When the study started (February 2012), service users assisted in developing the protocol and study information materials through feedback to the group representatives (HC and WR). Once ethics approval had been obtained (May 2012), the group representatives made the introductions between the service users and the fieldwork researchers. Researchers HM and GT attended 15 meetings between May 2012 and September 2013 to engage the service users by discussing study progress, gaining feedback and thus informing the research strategy. Reflecting on researcher perspectives is important and so reflexive diaries were kept and researcher observations were shared with the research team throughout.

Negotiating meeting frequency and types of visits

In Aberdeen, initial contact was made through the health visitor (WR); however, the researcher was able to exchange contact details with group members at her second visit and so the gatekeeper role shifted to the group’s treasurer, who shared it with one of the grandmothers who was responsible for organising meetings and events. This allowed for much less formal communication, good rapport building and thus intermittently ‘dropping in’ for informal lunch visits, as well as the more directed research sessions. The group met at a facility within walking distance of the university (approximately 10 minutes); this made visits easy to schedule and undertake. These circumstances were also fortunate in that, a few months into the project (September 2012), WR was relocated by her NHS employer and her replacement assumed a less prominent role within the group. In fact, after some further months (February/March 2013), health visitor support for the group was withdrawn altogether. However, this did not adversely affect engaging these service users and possibly strengthened a direct partnership between the researcher (HM) and the group. By contrast, in Blackpool, contact was initiated and continued through the community centre worker and therefore only formal research meetings were undertaken. The distance of travel for the researcher (18 miles by car) was a barrier to informal contacts. For both groups, long school holidays coincided with crucial research stages and were thus inopportune for our research schedule. For example, the final qualitative data analysis overlapped with the school summer holidays, although we were able to conduct an exercise involving our key qualitative findings with both groups (September 2013).

The Aberdeen group used Facebook (see www.facebook.com) as its primary communication mode; however, despite gaining ethical approval, the research team decided against engaging with the group using this social medium. Several concerns were discussed around maintaining appropriate boundaries, the best use of researcher time and a lack of published guidance. Using e-mail/telephone/text only did not appear to prevent or diminish engagement. Informal contacts and balancing communication traffic presented a challenge to the researcher (HM), who was confronted with some pressure for full group membership, such as being proactively invited to additional/extra activities: to assist at fundraising events and fêtes, to wear fancy dress to group seasonal parties and to attend mothers’ social evenings. The requirements of a researcher to remain objective and maintain a distance compared with a more anthropological approach to fieldwork were discussed with the research team. Of the six informal social events that took place during the course of this research, HM attended three to show goodwill. Researcher participation can therefore be considered as ‘active’101 as HM was able to gain insight into the cultural codes and rules for behaviour of the group.

Including service-user perspectives

Participatory approaches were employed and initially consisted of building good rapports with both groups by attending their meetings, informally socialising and observing in an unobtrusive manner to acknowledge the researchers’ roles as visitors to an established group. 101 This began with standing passively on the periphery, waiting to be invited in, eventually sitting comfortably on the sofas and even on the floor, and led to more active engagement, such as being asked to console crying babies. Once rapports were established, the researchers initiated requests for more directed sessions to gather feedback and data on important stages in the study.

Directed data-gathering sessions with the service users occurred after information was provided in advance and on an agreed date. HM and GT sought informed consent to audio record discussions. Researchers involved the groups in the design, editing and presentation of intervention vignettes derived from a diverse sample of promising studies identified in the systematic reviews (see Appendix 2). The study vignettes were then used in interviews and focus groups to gain participants’ perspectives of incentive interventions and to assist with shortlisting promising incentive strategies. Pilot focus groups were undertaken to contribute to interview topic guide refinement. We were able to generate data that contributed to the qualitative analysis from five of those sessions [two individual interviews and three focus groups (see Chapter 6)]. In addition, group members piloted the DCE as the majority had a smoking history, constructed their own intervention ‘ladders’ using the logic model that we developed (see Chapter 6), commented on the lay summary and advised on future dissemination of study findings. Our co-applicant groups also assisted us in identifying and actively engaging ‘hard-to-reach’ women, who seldom access health services and participate in PPI initiatives even less often. Although they contributed to the development work leading to the shortlist of incentive strategies asked about in the MORI survey of the general public (see Chapter 7), a separate sample of independent general public participants was sought to pilot this to minimise the ‘group think’ that may occur through repeated discussion of a topic. 102

Collaborative approaches to incorporating service-user perspectives

The Aberdeen group environment, cohesiveness and set structure suited a directed format, that is, focused activities, to address research issues. Women in this group were used to undertaking training (e.g. food preparation, child first aid) and had regularly welcomed external speakers in the second hour of their weekly 2-hour meetings. During this time, women sat on comfortable sofas around a large coffee table or leaned in over the kitchen counter if they were making the tea/coffee. Meanwhile, their children were occupied in the crèche, which meant that the researcher (HM) could interact with women informally as well as undertaking researcher-led whole-group data collection sessions, with minimal issues around audio recording and using interactive materials, such as the intervention vignettes. Although computing/projector facilities were not available, whiteboard and printed copies were used and worked fairly well. The vignettes helped the group to become focused and positively engaged with the study in a more concrete and tangible way. On one occasion, however, internal politics caused tension because the session was led by the organiser, who was upset that day and prevented any research activities taking place because she was ‘telling off’ members of the group about rota duties and fulfilling their responsibilities (e.g. washing up and cleaning the carpets).

In Blackpool, the group’s drop-in format presented the researcher (GT) with the practical challenge of having to move between members to involve them in ‘chance encounters’. Although the service users were keen to participate and engage in discussions, group membership was discontinuous and so GT was unable to engage with the same parents at each visit. Despite these limitations, a number of women were involved in various activities on more than one occasion, and an advantage was that a wider number of service users were engaged in total (n = 12) compared with the Aberdeen group (n = 8). Further challenges were faced when trying to engage parents and carers in meaningful conversations when their infants/children were in attendance. However, during piloting of the DCE, additional support was provided by children’s centre staff to ‘supervise’ the children to enable close reading and discussion of the online format of this tool. Furthermore, although the noise levels (children playing, shouting, etc.) compromised the digital recordings of the discussions, GT also kept handwritten notes and transcribed the session recordings herself as soon as possible to facilitate accurate and comprehensive data documentation.

Relationship with the wider research team

Engaging the mother-and-baby groups with the wider research team was problematic. Inviting and encouraging group members to attend regular weekly or formal research team meetings at the university (e.g. grant holders’ meetings) proved impossible at both sites. The health visitor group facilitator from Aberdeen (WR) attended one meeting (September 2012) and suggested that the women would have been very nervous had any of them attended because of the formality. Her successor was invited to a subsequent meeting and expressed an interest in the associated paperwork, which was e-mailed to her; however, she did not attend. It was not possible for Blackpool group members to attend because of the distance to travel and they did not accept an invitation to attend by telephone.

Other members of the research team were invited to join HM and GT for visits to the mother-and-baby groups. Highly sceptical comments such as ‘You’ll be lucky!’ were made by academic colleagues and administrative staff when considering the potential take-up of such an invitation. PH accepted on two occasions and attended an Aberdeen mother-and-baby group Christmas ‘drop-in’ (December 2012) and observed and took notes while HM facilitated a ‘ladders’ session (September 2013) (see Substantive contributions, Ladders). NC similarly observed and took notes while GT facilitated a ‘ladders’ session in Blackpool (September 2013).

The importance for the groups of researchers going into their territory to gain PPI input was evident. They took pride in hosting researcher visits, for example enthusiastically offering refreshments and home baking (Aberdeen) or computer facilities for directed sessions (Blackpool). In addition, researchers witnessed conversations that are seldom encountered in formal health or research settings. For example, some group members were adversarial towards health professionals, one group described complex and disruptive domestic relationships and highly charged personal opinions were expressed about health behaviours (smoking during pregnancy) and on one occasion someone left the room. In the presence of health service staff, such accounts would have been unlikely.

Substantive contributions

Designing participant materials

Through the group representatives as mediators, service users helped us to rephrase several sections of the information sheet for both readability and acceptability before gaining ethics approval and undertaking the primary qualitative research (Figure 3).

FIGURE 3.

Comparison of the patient information leaflet before and following service-user input.

We revised the document to take account of the target participants’ style preferences, language and social and cultural contexts. In particular, they drew our attention to their unfamiliarity with the term ‘cessation’ and the predominance of formula feeding, either from their own personal experience or in their immediate family and social networks. Thus, they suggested that ‘help women to stop smoking’ and ‘try breastfeeding’ were more appropriate phrases, and they also pointed out that ‘encourage’ is persuasive, ‘smoking cessation’ is too technical and the word ‘breastfeeding’ alone implies a certainty. The last of these was a very important change as we noted that a mother in one of the groups disclosed that she was providing exclusive breast milk to her infant (on her information form) but pretended that she was formula feeding to other group members, using bottles of expressed breast milk that she allowed them to assume was formula by talking about preparing it. Breastfeeding was not a social norm and this woman anticipated that it could possibly be considered unacceptable by some.

Research questions

1. What is the evidence for the effectiveness of incentive interventions delivered within or outside the NHS to (a) individuals and/or their families or (b) organisations that seek to increase and sustain (i) smoking cessation during pregnancy and/or within the first 6 months following birth and (ii) breastfeeding in the first 6 months following birth?

2. What is the evidence for effective incentive delivery processes (including acceptability, mechanisms of action, barriers/facilitators and intrinsic or extrinsic motivators to behavioural change) for recipients and/or providers to increase and sustain (i) smoking cessation during pregnancy and/or within the first 6 months following birth and (ii) breastfeeding in the first 6 months following birth?

Objectives

1. To systematically review incentive strategies for (i) smoking cessation in pregnancy and/or within the first 6 months following birth and (ii) breastfeeding within the first 6 months following birth to gather evidence to inform the design of an incentive intervention trial and to address the overall study objectives detailed in Chapter 1.

2. To integrate the quantitative and qualitative evidence on incentives for (i) smoking cessation in pregnancy and up to 6 months after birth and (ii) breastfeeding up to 6 months after birth.

For each of the two behaviours of interest – smoking cessation and breastfeeding – an integrated mixed-methods systematic review of effectiveness and delivery processes was undertaken. The effectiveness review was informed by the principles of Cochrane guidance on the conduct of systematic reviews,117 with particular reference to public health intervention guidance. 118 The process review was guided by the nature of the studies identified and the overall study research questions. The process review includes any data that meet the study inclusion criteria but are not included in the effectiveness review. The protocols were registered with PROSPERO (reference no. CRD42012001980) (see Appendix 6).

Identification of studies

Extensive electronic searches were conducted to identify reports of published, unpublished and ongoing studies that report data on the benefits of incentives for smoking cessation during pregnancy and/or within the first 6 months after birth and initiating and/or sustaining breastfeeding within the first 6 months after birth. The search strategy was designed to be highly sensitive and include appropriate subject headings and text-word terms. Full details of the search strategies used are reported in Appendix 7.

The databases searched were MEDLINE (1946 to February 2012), EMBASE (1974 to February 2012), Cumulative Index to Nursing and Allied Health Literature (CINAHL) (1981 to February 2012), Science Citation Index (1981 to February 2012), Social Science Citation Index (1981 to February 2012), Applied Social Sciences Index and Abstracts (1987 to February 2012), PsycINFO (1967 to February 2012) and Trials Register of Promoting Health Interventions (February 2012). In addition, the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library, Issue 2, 2012) was searched for additional reports of randomised controlled trials (RCTs) and the Health Management Information Consortium database (1979 to February 2012) and the NHS Economic Evaluation Database (February 2012) were searched for economic literature. The Cochrane Database of Systematic Reviews (CDSR) (The Cochrane Library, Issue 2, 2012), Database of Abstracts of Reviews of Effects (DARE) (February 2012) and the HTA database (February 2012) were searched for relevant systematic reviews and reports. Relevant websites, including the WHO website (see www.who.int) and those of the United Nations Children’s Fund (UNICEF) (see www.unicef.org), the King’s Fund (see www.kingsfund.org.uk/), the Health Foundation (see www.health.org.uk), the Joanna Briggs Institute (see www.joannabriggs.edu.au), the Picker Institute (see www.pickereurope.org) and the Midwives Information and Resource Service (MIDIRS) (see www.midirs.org) were also searched. Co-applicants also used their networks to identify relevant studies.

Definition of an incentive

Incentives included financial (positive or negative) and non-financial but tangible incentives or rewards. Tangible incentives were defined as either free or reduced-cost items that have a monetary value (e.g. refreshments or baby products) or services with a monetary value (e.g. childcare or ironing). These incentives could be delivered directly or indirectly at local, regional or national level by health-care or other community and/or commercial providers. This excluded supportive, persuasive or motivational relationships with service providers or peers or educational materials such as information leaflets or non-commercial DVDs that reinforce verbal persuasive advice to stop smoking or to breastfeed.

Inclusion and exclusion criteria

The reviews included English-language studies of any study design (RCT, quasi-RCT, non-randomised interventions, mixed-method designs and qualitative studies) that met the definition of incentives and which were from a developed country as defined by the United Nations. 119 Grey literature was identified according to the Luxembourg definition of multiple document types produced by all levels of government, academia, business and organisations in electronic and print formats not controlled by commercial publishing, that is, in which publishing is not the primary activity of the producing body. 120

The purpose of the incentive could be (1) a behaviour change outcome or (2) participation in either an intervention or monitoring of the behaviour. We were interested in how incentives/rewards as BCTs were being used by researchers in intervention studies either alone or in combination with other BCTs.

The population of interest was women who were pregnant or those who had given birth within 6 months at the time of the intervention (either including or not including other family members) and/or health-care (e.g. NHS or other health-care companies), community (e.g. local authority or voluntary sector) or commercial (e.g. pharmacy) providers and/or stakeholders at local, regional or national level. Interventions could benefit one or more than one of these groups. The review concentrated on interventions specifically targeted at the population of interest and did not include studies analysing the population of interest as a subgroup within a wider population at whom the intervention was aimed.

Comparators under consideration included any alternative treatment arms to those of the interventions that represented either practices similar to those currently being provided in countries meeting the inclusion criteria for the review or possible alternative interventions that are realistic options that may be relevant to health-care and/or community providers in these countries, particularly (from a UK perspective) stakeholders within the NHS. Single-group before-and-after studies were also included.

Our primary outcomes were smoking cessation rates and relapse rates and any breast milk and exclusive breast milk feeding rates. All studies for the effectiveness review were required to provide data for at least one of these outcomes. In addition, when possible, we collected secondary outcomes related to the delivery of the intervention or outcomes related to the characteristics of the incentive mechanism. We were particularly interested in:

• the context of the intervention:

  • the behaviour change theories informing the intervention

  • the setting within which the intervention was delivered

  • who the intervention was delivered by

  • how the intervention was delivered, for example in person, remote delivery, number of person contacts involved in delivery

• for outcomes related to the characteristics of the incentive mechanism:

  • who is being incentivised (e.g. individuals, their families and/or provider organisations)

  • other incentive characteristics (e.g. size, duration, timing, delivery, ‘currency’, that is, financial or tangible non-financial)

  • the fidelity (compliance with the protocol) of the intervention and/or the incentive/reward mechanism

  • details of whether the incentive is contingent on something and if so what it is contingent on (e.g. abstinence from smoking, initiation or duration of breastfeeding, attendance at clinic, monitoring to validate primary outcomes either biochemically for smoking cessation or self-report, achieving other programme components)

  • differential effects, for example health inequalities evident.

The gold standard for biochemical validation of smoking outcomes is considered to be the level of cotinine in blood or saliva. Cotinine measurements are more reliable than exhaled carbon monoxide (CO) levels as levels of the latter can be undetectable within < 24 hours. 121 The half-life of cotinine in pregnant women between 16 and 40 weeks’ gestation is 8.8 hours compared with 16 hours for non-pregnant women and clearance to non-detectability takes place in about 44 hours. 122 Cotinine-level cut-offs vary, largely dependent on the levels of sensitivity and specificity chosen. The plasma level cut-off used is often 13.7 ng/ml but more recent studies in the pregnant population suggest that because of increased clearance of nicotine and cotinine during pregnancy a lower cut-off of 11.5 ng/ml may provide greater accuracy. 123 Saliva levels are thought to mirror plasma levels, with a cut-off of 13 ng/ml quoted for pregnant smokers in a recent study. 124 Urine cotinine levels are in theory more problematic in that the cotinine level should be corrected for dilution by measuring creatinine in the urine. In practice, the dilution effect may not be important and laboratories differ, with a cut-off of around 44.7 ng/ml. 125 Bedside assays for cotinine using urine and saliva are being developed. 126

Studies that reported secondary outcomes without also reporting at least one of the primary outcomes were excluded, with the exception of incentive studies that were not linked to a specific intervention but focused on experiences/perceptions of incentives for smoking around childbirth or breastfeeding (see Data extraction strategy). When interventions included both an incentive/reward behaviour change technique (IRBCT) component and general BCT components (e.g. support/motivational counselling), these studies were included regardless of the relative size of each component. The studies included in this evidence synthesis chapter are listed in Appendices 8 (smoking cessation) and 9 (breastfeeding).

Data extraction strategy

Abstracts identified by the search strategy were screened by two reviewers (JH and HM for smoking cessation and KR and HM or KR and PH for breastfeeding). Full-text papers of potentially relevant reports were obtained and independently assessed by two reviewers for inclusion in the effectiveness and process reviews (JH and KR for smoking cessation and KR and one of GT, NC or DT for breastfeeding). Qualitative and mixed-methods papers that met inclusion criteria only for the process review were independently assessed by two reviewers (HM and PH). Differences of opinion were resolved through review team discussion. Authors of individual papers were contacted for additional information when it was required to answer the research questions. The grey literature was assessed by two reviewers – for smoking cessation studies and KR and HM for breastfeeding studies – for inclusion in either the effectiveness or the process reviews; however, following team discussion it was considered more appropriate to use these data to inform triangulation of the findings for the entire study and to inform qualitative data collection. The grey literature is summarised in Appendix 10 and is referred to throughout the report.

As no purely qualitative studies were identified, the original analysis protocol was modified to maximise the information available on incentive delivery processes reported in the included studies. Studies for process data extraction were grouped into two categories:

1. Qualitative and mixed-method studies linked to identified RCTs and intervention studies, including any descriptive data reported as part of papers reporting quantitative outcomes. This is recommended in critical interpretive synthesis127 and in a review of smoking cessation interventions in pregnancy to ensure that findings are relevant for service users and practitioners. 128

2. Qualitative or descriptive studies not linked to any specific intervention. This included studies drawing on the experiences and perceptions of incentives for smoking cessation and breastfeeding around childbirth.

A single electronic mixed-methods data extraction form (see Appendix 11) for use in Microsoft Word (Microsoft Office Professional Plus 2010, Microsoft Corporation, Redmond, WA, USA) was developed and agreed through review team discussion, informed by the Cochrane public health guidance. 118 A single data extraction form was considered to be possible for the dual purposes of conducting a hypothesis-testing (effectiveness) review and a hypothesis-generating (process) review.

One reviewer extracted quantitative data from the included full-text papers first (JH for smoking cessation and KR, VHM or SJ for breastfeeding) and a second reviewer (KR for smoking cessation and KR, VHM, SJ, LJ or NC for breastfeeding) – depending on who originally extracted data from each study – checked the data extraction. The data extraction form was then passed to the process review team where one reviewer (HM) extracted hypothesis-generating data relating to the delivery of and perceptions about incentive interventions and a second reviewer (GT or PH) checked the data extraction. For the studies reporting qualitative data, identified themes were integrated with the mixed-methods review.

Behaviour change techniques

Intervention content was characterised by extracting BCTs for studies in which an intervention was delivered. Studies providing incentives to providers of care were excluded from BCT extraction. BCTs were extracted in two ways: (1) general BCTs (not involving incentives/rewards) were extracted to characterise the range of general BCTs employed and (2) IRBCTs were extracted for an in-depth examination of these components. General BCTs were extracted using behaviour-specific taxonomies17,129 to utilise the most appropriate frameworks to characterise intervention content. IRBCTs were excluded from the general BCT extraction as these were covered in the in-depth examination. In-depth examinations of IRBCTs were based on extractions using the recent BCT-v1,17 which provides the most comprehensive cover of incentive/reward-related techniques (i.e. taxonomy section ‘reward and punishment’ 10.1–10.11). The IRBCTs were coded in a modified taxonomy created for this study derived from the literature review (Table 2). This is based on the four dimensions inherent in the ‘reward and punishment’ BCTs from BCT-v1,17 namely contingency target (i.e. health behaviour such as self-reported smoking abstinence, preparatory behaviour such as intervention participation and engagement, or outcomes such as CO measurement), actor (i.e. participant, health-care professional, social other), content (i.e. material, social, self, non-specific) and type/awareness (i.e. aware of reward in advance and therefore acting as an incentive; not aware of reward; unclear). Social rewards were maintained for completeness, although they did not meet our inclusion criteria for the review. All BCTs were coded using a conservative approach, which required explicit written evidence for the use of a technique to avoid subjective inferences. Data were not extracted for the control arms because of poor-quality reporting, particularly of usual care.

Smoking cessation studies

Behaviour change techniques were extracted using a smoking cessation-specific taxonomy129 that includes 41 smoking-relevant BCTs (the BCTs ‘provide rewards contingent on successfully stopping smoking’ and ‘provide rewards contingent on effort or progress’ were not rated as these were examined in the in-depth section based on the BCT-v1 taxonomy). BCTs were rated for the target behaviour of stopping smoking in the target population of pregnant women and/or those who had given birth within the last 6 months. The definition of the BCT ‘measure CO’ was extended to also include objective verification of smoking status through saliva samples. Two reviewers (HM and SUD) initially extracted BCTs for three studies that had been excluded from the review but which were very similar to included studies to pilot the extraction procedures and documents to obtain good inter-rater reliability. A consistent extraction and coding strategy was agreed following comparison and discussion. The reviewers then independently extracted BCTs from the included papers. Discrepancies in BCT codings were compared and resolved by a third senior reviewer (FFS).

Breastfeeding studies

A similar approach was used for the breastfeeding papers; however, in the absence of a behaviour-specific taxonomy for breastfeeding, the generic BCT-v1 taxonomy17 was applied. BCTs were rated for the target behaviour of breastfeeding in the target population of woman who were pregnant or who had recently given birth. The definition of the BCT ‘adding objects to environment’ was refined to explicitly exclude information-based materials such as leaflets and booklets as these were judged to be qualitatively different from the provision of objects such as breast pumps.

Quality assessment strategy

Two reviewers [JH and KR for smoking cessation and KR or VHM and SJ (acknowledgement) for breastfeeding] independently appraised the quality of studies included in the effectiveness review using the Cochrane Collaboration’s risk of bias tool130 for RCTs and/or (depending on the design of identified studies) an 18-question checklist for non-randomised studies and case series that was adapted from several sources including the Centre for Reviews and Dissemination’s guidance for those carrying out or commissioning reviews,131 Verhagen and colleagues,132 Downs and Black133 and the Graphic Appraisal Tool for Epidemiology (GATE). 134 This tool assesses bias and generalisability, sample definition and selection, description of the intervention, outcome assessment, adequacy of follow-up and performance of the analysis. The checklist was developed through the NICE Review Body for Interventional Procedures (ReBIP), a joint venture between the Health Services Research Unit at the University of Aberdeen and Health Services Research at the University of Sheffield for the NICE Interventional Procedures Programme. Any differences of opinion between independent reviewers were resolved through discussion with a third member of the team.

The quality of papers reporting qualitative data was independently assessed by two qualitative researchers HM or GT or NC; for one study authored by members of the research team, SM and VS (see Acknowledgements) assessed quality using a modified version of the Critical Appraisal Skills Programme (CASP)135 (see Appendix 12). Multimethod studies included in the effectiveness review were also assessed for risk of bias as described above. When consensus could not be reached, wider discussion with the research team took place to reach agreement. We also applied the Mixed Methods Assessment Tool (MMAT)136 to studies reporting quantitative and qualitative data, including survey studies; however, this was abandoned as it did not contribute additional useful information and it was insufficiently sensitive to detect quality issues with the surveys.

Data analysis

For the quantitative results we report means or changes in means or proportions between groups. For data that were reported only as figures (i.e. without information in the text of the article), data have been digitised using Engauge software (Digitizer version 4.1, see http://sourceforge.net/projects/digitizer/files/Engauge%20Digitizer/) to maximise the available information and report data as precisely as possible. For relevant outcomes, meta-analysis was carried out when appropriate to estimate a summary measure of effect, using Stata 13 (StataCorp LP, College Station, TX, USA), by calculating relative risk (RR) and risk difference (RD) statistics, using random-effects methods to account for heterogeneity. Inter-rater agreement was assessed using Cohen’s kappa137 if BCTs were identified in five or more instances. For studies not considered suitable for meta-analysis but in which primary outcomes were reported for intervention and control groups, RRs were calculated from the outcome data reported.

A narrative synthesis was conducted for the incentive process review. Process review analysis involved the qualitative research team (NC, PH, HM, GT) familiarising themselves with the studies and discussing the data extracted, developing initial descriptive themes and finally developing higher-order analytical and interpretive themes and concepts. Service-users’ perspectives were incorporated into this process through discussion with our co-applicant mother-and-baby groups (see Chapter 2). This cyclical and iterative process was used to identify the promising ‘ingredients’, processes and temporal aspects of incentive interventions most likely to contribute to effectiveness for smoking cessation in pregnancy and breastfeeding, including situation, context and environment themes.

We now present the results of the smoking cessation review. This is followed by the results of the breastfeeding review and the results of studies that incentivised providers.

Number of studies identified

We identified 1469 records from the primary searches for this review. After title and abstract screening, 1161 studies were considered not to be relevant and were excluded, leaving 215 studies for full-text assessment and a further 93 studies that were requested for background information only. An additional seven studies were identified by searching reference lists and from research team knowledge. In total, the full texts of 222 studies were therefore screened. In total, 23 studies (34 reports25,68,104,105,108–111,113,138–162) met our inclusion criteria. A flow diagram of the screening process is provided in Figure 4.

FIGURE 4.

Flow chart outlining the screening process for the review of incentives for smoking cessation in pregnancy and within 6 months of giving birth.

Included and excluded studies

Of the 23 studies included, 20 provided incentives to childbearing women. 104,105,108–111,113,138–150 One study was included in the process review only as it was a survey of acceptability and feasibility and did not provide incentives to women. 151 One study152 provided reimbursement to organisations providing care and one study included a question in a survey about reimbursement. 153 The surveys were included to inform the primary survey research to investigate the acceptability of incentives reported in Chapter 7. Studies providing incentives to organisations for either smoking cessation or breastfeeding are reported separately later in this chapter. The reasons for exclusion of assessed full-text papers are given in Figure 4.

There were reports in which it was unclear whether a study met our definition of an incentive. Decisions to include as an incentive study were made for the following:

1. A study by Ussher and colleagues150 in which free expert-led exercise sessions with free gym equipment and an exercise DVD were provided. As free expert tuition, a DVD and use of gym equipment has a cost, this met our criteria for an incentive intervention.

2. Multifaceted programmes providing incentives to women as part of usual care, for example the Women, Infants and Children (WIC) programme (see www.fns.usda.gov/wic/). The WIC programme provides federal grants to states for supplemental foods, health-care referrals and nutrition education for low-income pregnant, breastfeeding and non-breastfeeding post partum women and infants and children aged up to 5 years who are found to be at nutritional risk. It was decided to include only studies in which the aim was to investigate an incentive intervention according to our study definition rather than to include all WIC studies.

Decisions to exclude as an incentive study were made for the following:

1. Studies in which it was unclear whether an item provided was intended as an incentive or as a treatment or therapy that was being subsidised as part of the research were excluded. For example, NRT is provided free to NHS patients within the UK; however, in other countries this may not be the case. In such countries a pregnant smoker could be incentivised to stop smoking if they had free access to NRT.

Quality of the included studies

Details of the questions used to establish the quality of all included studies presenting quantitative outcome data are available in Appendix 13, where the individual assessment results for each question by study are reported.

Information is provided about each individual aspect of quality across all studies (Figure 5) and across all aspects of quality for each individual study (Figure 6). For both figures, answers of ‘yes’ represented the maximum possible level of quality. Therefore, in considering Figure 5, it can be seen that most studies (> 70%) had clearly described their inclusion and exclusion criteria, had undertaken data collection prospectively, had clearly defined their interventions and comparators, had followed up participants for long enough to detect important effects on outcomes and had considered and reported all of the important outcomes.

FIGURE 5.

Quality assessment results by quality appraisal question for all studies included in the review of incentives for smoking cessation in pregnancy.

FIGURE 6.

Quality assessment results by study for all aspects of quality for the review of incentives for smoking cessation in pregnancy.

For some studies, details regarding the study population were absent and so it was less clear that participants were from a representative sample of a relevant patient population or that groups were comparable on demographic characteristics, although in some studies there was no comparator group and so it was not possible to answer this question or those regarding whether length of follow-up was similar between groups and whether groups were treated identically other than for the named intervention. For most studies (i.e. > 50% of studies) objective (valid and reliable) outcome measures had been used (e.g. biochemical validation of cessation).

More problematic aspects of study quality were the lack of clarity regarding the selection of patients and allocation concealment for studies reporting to be randomised. The effect of loss to follow-up on smoking cessation results was also unclear. Consistent with the Russell standard for outcome criteria in smoking cessation trials,163 we assumed for the purposes of our analysis that loss to follow-up would result in smoking relapse.

With regard to Figure 6, studies with the lowest proportion of ‘yes’ answers were those with an atypical study design. For example, the studies by Edwards and colleagues142 and Morgan and colleagues148 considered the effect of the interventions at a population level and the studies by Mantzari and colleagues,109 Nichter and colleagues110 and Ripley-Moffitt and colleagues111 were predominantly qualitative with additional quantitative results mentioned. In addition, several studies, including those by Cluss and colleagues140 and Ussher and colleagues,150 did not have comparator groups and therefore some of the questions considering the appropriateness of how groups were selected and treated did not apply.

The overall quality of the qualitative content of the four included qualitative and mixed-methods studies,109–111,113 assessed using CASP,135 was high and the detail is presented in Table 3.

Characteristics of the included studies

Studies were classified from A to E according to incentive and study design. This classification is presented in Table 4.

The main characteristics of the 20 included studies that report incentives for smoking cessation in childbearing women are described in Table 5. Detailed characteristics of the incentive/reward are described in Table 6; the setting, provider and intensity of incentive delivery are summarised in Table 7 and the chronological patient journey through the intervention showing the timing of intervention components and BCTs offered to study participants is shown in Table 8.

Most studies originated from the USA and they fell into two main categories: (1) contingent incentives for smoking cessation or (2) non-contingent incentives for participation in a smoking cessation programme and providing outcome data. This latter category included incentives for participation in another intervention component in the study or research to understand participants’ smoking behaviour. Within these two broad categories, a range of comparisons was present. Three studies were considered to have two incentive intervention arms. 138,139,147 The studies by Albrecht and colleagues138 and McBride and colleagues147 are B2 studies; both intervention arms involved incentives and these are compared with a comparator arm (usual care). The study by Cinciripini and colleagues139 is a category D study; both study arms received incentives and differed only in terms of another non-incentive component.

Intervention content and format

Studies providing multicomponent incentive interventions to aid smoking cessation among pregnant women are summarised in Table 6. All incentive interventions involved interactions with incentive providers or other professionals, either directly or indirectly supporting smoking cessation, and we refer to these as the general BCT components of the intervention. However, communication at the time of delivering the incentive is under-reported and no study mentions observations or recordings of interactions during intervention delivery. The proportion of the intervention that can be attributed to the incentive and to the BCT component varied between studies and was sometimes difficult to assess. In category A studies105,109,141,144 the incentive was the dominant component. In some studies the primary study aim was to investigate separate BCT components, with the incentive of secondary importance, for example to improve recruitment, intervention uptake or reduce attrition. For category B–D studies, women were invited to participate in a range of activities to obtain the incentives, some of which were reported as usual care plus additional components110,111,138 or an enhanced programme of care108,145–149 or involved the implementation of a new protocol. 104,113,139,140,142,143,150 These programme components involved either one or a combination of advice, counselling, group sessions, written materials and motivational strategies and, in one study, a programme of supervised exercise. 150

Incentive/reward-related elements

When extracting the IRBCT data, incentive and reward elements were identified reliably (kappa = 0.87). Out of the 20 studies, eight104,109–111,139,142,145,148 used multiple incentives (see Table 6) within their interventions, which varied in the four assessed incentive elements: material, social, self-selected and non-specific (see Table 2).

All studies included material incentives/rewards as this was a criterion for study inclusion. Two interventions specifically mentioned including social incentive/reward components such as congratulatory cards for being abstinent148 and health-care professional praise146 as well as the material incentive. In addition, one intervention offered supervised exercise classes. 150

Incentive categories emerged through analysis of the description of the incentives and the context in which they were delivered:

• Earned rewards, for example vouchers contingent on biochemical validation of quit status as in the studies by Heil and colleagues,105 Higgins and colleagues144 and Donatelle and colleagues. 141

• Prizes or awards – these could be either contingent on programme participation as in the study by Pbert and colleagues149 or contingent on stopping smoking. They were either predictable or unpredictable as in the raffles provided in the study by Gulliver and colleagues. 104

• Gifts that were not contingent on either programme participation or quitting, with the aim of encouragement or increasing goodwill towards the programme deliverers or research team, as in the study by Edwards and colleagues. 142

• Compensatory, for example voucher-based reinforcement therapy (VBRT). This was for participating in BCT programmes that might be demanding in time and commitment, as in the study by Cinciripini and colleagues. 139

Further analysis of the meanings of these emerging incentive categories is provided in Chapter 6, where qualitative data contribute to an incentive typology.

General behaviour change techniques

One study152 was excluded from BCT data extraction as the intervention was rolled out at the health system level rather than at the individual level and was thus not in line with the target behaviour.

When extracting the general BCT data, inter-rater agreement was acceptable, with a kappa of 0.78, ranging from 0.66 (‘offer/direct towards appropriate written materials’) to 0.94 (‘facilitate goal-setting’). Explicit reporting of BCTs underpinning interventions varied and this was often unclear or implicit. Altogether, 32 study arms were rated for the inclusion of 41 BCTs, including nine control/usual care arms and 23 intervention arms. Intervention arms were identified to use an average of 6.62 [standard deviation (SD) 3.1] BCTs, ranging from one109 to 12,145 emphasising that incentive interventions are complex and consist of various potentially active ingredients. Control/usual care arms used an average of 1.11 (SD 0.9) BCTs, ranging from zero109,143 to three. 138

Out of 41 BCTs rated, study arms were identified to have used 30 (Figure 7), with the BCTs used most often being ‘providing information on the consequences of smoking cessation’ (n = 24), ‘measuring CO’ (n = 18), which mostly took the form of objectively validating smoking status within intervention encounters, and ‘facilitating goal-setting’ (n = 15), which in most studies took the form of setting a formal quit date.

FIGURE 7.

General BCTs used in all intervention arms in the studies included in the review of incentives for smoking cessation in pregnancy.

In addition to the BCTs, six studies were also rated as being engaged in the WIC programme105,141–143,145,149 and one study was rated as being partially engaged in the WIC programme. 110 BCTs used in the WIC programme were recorded separately without inferring the presence of specific BCTs given the variation in WIC programme implementation.

The majority of individuals targeted to receive the incentives/rewards were judged to have been aware of these components (n = 12); of the remaining rewards, some were judged as not being known to participants (n = 5) and it was unclear whether participants were aware of the others (n = 5).

Setting and provider

Included studies were mostly from the USA, with three109,113,148 set in the UK (see Table 5). The studies varied in terms of their funding sources. One was funded by a research council108 and several were funded by private trusts. 104,109–111,142,151 One study was jointly funded by a private trust and a national health provider142 and the remainder were funded by national health or government institutes,113,138–140,143,144,146–150 with two receiving extra financial support from the WIC programme. 105,145 In addition, two explicitly relied on donations from local businesses to provide the incentives,104,108 which was framed as helping to reinforce the message of community support for smoking cessation through drawing women’s attention to it, as well as being cost-saving and making the programme and its continuation more attractive and sustainable to principal funders. The authors’ disciplinary backgrounds were also assorted and often involved multidisciplinary collaborations among research teams. A number of studies had a behavioural/psychological background108,109,139,147,149 and many involved public health/health researchers,113,141,142,146,150 with one overlapping both areas. 151 A number of studies were led by psychiatric specialists105,140,144 and two had other origins: education (paired with midwifery)148 and nursing. 138 Several had a cancer-specific derivation145–147 and two were approached by substance abuse/addiction experts. 104,111 One study was led by anthropologists who used ethnographic methods to understand participant perspectives. 110

The intervention settings and providers varied across the included studies (Table 7). A number of the studies were WIC based and were delivered in WIC clinics by WIC staff as an enhancement of usual care/smoking in pregnancy interventions. 105,141–143,145,149 One study recruited women through the WIC programme but women were interviewed in their own homes. 110 Similarly, other studies also involved improved or more intensive programmes designed for smoking cessation in pregnancy delivered alongside or through health services (e.g. the NHS) where contact with pregnant women was already taking place. 109,148 The others were mostly delivered in health-care settings but involved attending a non-standard clinic. 104,111,113,138–140,146,150 One intervention was delivered through an army hospital147 and others specifically recruited pregnant women to their studies and delivered the interventions in other settings, such as a university-based clinic,111,144 a teaching hospital108 and in schools. 138 Although most of the programmes were delivered in clinics by health professionals or research staff, some also involved trained workers carrying out home visits/offering telephone support,110,113,148 a helpline109,143 or telephone calls. 141,149 Details were often provided about the setting and staff providing the counselling-based components of the intervention, but little information was given regarding the process of providing incentives to participants. Some differentiation was made between the staff delivering the intervention, although in most cases this was carried out by usual-care health professionals: WIC staff,142,143 doctors and midwives,108 clinical psychologists,104 nurses138,146 and pharmacists. 113 The details reported were variable, but some noted when staff were specifically trained to deliver the intervention. 113,140,141,148 Study staff were also involved in delivering the intervention in many of the studies. 104,105,109–111,138,139,141,143–145,147,149 A number of the studies involved a community element whereby the smoking cessation message was explicitly reinforced either at routine clinic appointments in addition to intervention components108,140,146 or through community-based messages. 104

Intervention intensity and the participant journey

Studies provided multicomponent interventions to aid smoking cessation among pregnant women, with interventions (or even individual components within the intervention) provided at varying frequencies and time points. This is summarised in Tables 7 and 8. Intervention intensity refers to the number of contacts involved for patients from the start of the intervention until the final outcome data are collected; the study duration; and the length of individual encounters. The participant journey refers to a participant’s experience of the intervention components (time points and frequency) during the course of the intervention.

The time points at which women were recruited to the studies, when they first attended and when they received the first incentive varied and were often unclear. In particular, differentiation between contacts that were integrated with usual care and contacts that were separate from usual care was often unclear. In several studies it appears that recruitment could occur at any point during pregnancy. 104,109,113,140,142,144–146,148 When specific details were given, the earliest point at which women were reported to be recruited was at 4 weeks’ gestation138 and the latest was at 32 weeks’ gestation. 139 The incentive, or first incentive if more than one was given, was not necessarily provided at the point of recruitment; however, an incentive was given at initial intervention sessions in all studies with the exception of one. 143 In this study, the incentives (vouchers for nappy purchase) were not provided until the post partum period. In all of the other studies, incentives included items that were delivered during pregnancy. The overall number of contacts ranged from one to 36. In four of the studies, in addition to receiving the intervention, participants were also given usual care,108,138,144,147 whereas in three of the studies usual care was the control condition. 105,143,149 It might be assumed, however, that usual care was delivered in addition to the intervention in the remainder of the studies when this was not made clear, as well as for those in which it was the control condition. This, and the number of contacts that usual care might involve, was poorly reported throughout.

Some of the studies reported intensity in more detail with weekly attendance for the duration of the treatment and two involved daily monitoring for a period of 5 days. 105,144 Duration of treatment varied, with some studies delivering the programme components regularly over a short period of time, for example several, often consecutive, weeks,104,108,110,113,138–140,143,145,147–150 and some involving assessment at more spaced time points, for example at baseline, before delivery and postnatally. 111,141,142,146 Others were more varied, for example weekly to begin with and then after 3 months and 5 months. 109 End of treatment appears to have occurred before birth in some of the studies,138,140,150 although it is sometimes unclear. Six studies involved follow-up by way of an interview or assessment at a specified time point up to 6 months post partum. 105,139,143,144,147,149 In two studies contacts continued up to 1 year after birth,143,147 with one study providing monthly visits. 143 In another study an optional telephone helpline was available up to 1 year after birth. 113

Cinciripini and colleagues139 reported the mean duration of therapy sessions as 58 minutes (SD 10.1 minutes), with no significant differences in length of session between treatment groups, although in this study both treatment groups were incentivised to participate.

Ussher and colleagues150 reported the mean minutes of supervised exercise recorded for one of the studies they conducted that used exercise as the main component of the intervention, with supervised exercise increasing from 8.4 minutes (SD 3.5 minutes, n = 22) at the first visit to 19.9 minutes (SD 8.6 minutes, n = 16) on the first day of cessation, 23.3 minutes (SD 6.1 minutes, n = 6) 5 weeks after the first day of cessation and 18.5 minutes (SD 14.5 minutes, n = 6) 6 weeks after the first day of cessation.

Intervention fidelity/non-compliance

The fidelity of the intervention, in terms of whether the intervention was actually delivered or implemented in the way described in the protocol, was seldom reported. The intervention provided in the study by Gulliver and colleagues104 required participants to select a quit day; however, six of the 10 participants did not quit on their selected quit day.

Outcome measurement

In terms of incentive targets, 15 studies targeted preparatory behaviours (Figure 8). Preparatory behaviours targeted through incentives were intervention attendance and participation behaviours (n = 11), stating an intention to quit (n = 1), engaging in exercise (n = 1), counselling pregnant smokers (n = 1) or not specified (n = 3). Nine studies were identified as targeting only preparatory behaviours. 110,111,138–140,145,147,149,150 Six studies targeted preparatory behaviour in addition to outcomes (n = 4),105,141,142,144 health behaviour (n = 1)146 or outcomes and health behaviours simultaneously (n = 1)109 within the same intervention. Other studies targeted outcomes108,113,143 or health behaviours only. 148

FIGURE 8.

Combinations of contingency targets for incentive/rewards.

Outcome assessment varied depending on the study methodology used. Abstinence was point prevalence alone in 11 studies. 104,109,113,138,140–142,144,146,148,149 Seven studies attempted to quantify both point-prevalent and continuous abstinence. 105,108,139,143,145,147,150

Of the studies that had considered point-prevalent abstinence only, ten out of 11 (81.8%)104,109,113,138,140,141,144,146,148,149 had biochemically validated this. Seven studies104,109,113,138,140,144,148 had done so using CO levels, with the cut-offs ranging from ≤ 6 p.p.m. 144 to ≤ 10 p.p.m.. 104 The cut-off was unspecified in the studies by Morgan and colleagues148 and Radley and colleagues. 113 Two studies141,146 used thiocyanate to biochemically validate abstinence although only the study by Donatelle and colleagues141 specified the cut-off for this as 100 μg/ml. The studies by Donatelle and colleagues,141 Higgins and colleagues144 and Pbert and colleagues149 also used cotinine validation, with cut-offs ranging from 20 ng/ml to 80 ng/ml. The remaining study relied on self-reporting of smoking status. 142

Of the studies that used both point-prevalent and continuous abstinence, five out of seven (71.4%) had biochemically validated this. 105,108,139,143,150 Four studies105,139,143,150 had done so using CO levels, with the cut-offs ranging from < 4 p.p.m. 139 to < 8 p.p.m.,150 The studies by Cinciripini and colleagues139 and Heil and colleagues105 also used cotinine validation, with cut offs of < 15 ng/ml and ≤ 80 ng/ml respectively. The study by Walsh and colleagues108 used urinary cotinine validation, with a cut off of < 500 nmol/l. The remaining two studies did not use biochemical validation. 145,147

Of note, two studies had used biochemical validation methods for a different outcome, namely to quantify the deception rate, as it is possible to temporarily stop smoking to achieve a low CO level; however, this would be revealed by a raised cotinine level. 143,145 Gadomski and colleagues143 used salivary cotinine for this purpose (cut-off not specified) and Lillington and colleagues145 used a saliva cotinine cut-off of 20 ng/ml.

For the purposes of comparing data it was not always possible to utilise comparator data even if there had been a control group in the original study. For example, Gadomski and colleagues143 analysed data for an intervention with both gestational and post partum components. The provision of incentives applied to the post partum aspect of the intervention (vouchers for nappies); however, the authors did not have comparator data for this part of the intervention even though comparator data were available for the gestational part of the intervention (Anne Gadomski, Bassett Healthcare Research Institute, New York, May 2012, personal communication with JH).

In other cases it was not possible to derive outcome data for separate treatment groups. For example, the study by Albrecht and colleagues138 had three treatment arms comparing an intervention (Teen FreshStart), enhancement of the intervention with peer support (Teen FreshStart Plus Buddy) and a control group. However, results are reported comparing the enhanced intervention with combined data from the other two treatment arms; separate data for each of these arms are not available.

In two studies collecting population-level data had been attempted to consider the effectiveness of the interventions. 142,148 In the study by Edwards and colleagues142 data were gathered from forms that indicated whether or not a pregnant woman received a smoking cessation guide, which was provided only to women who expressed a desire to stop smoking (who were then offered counselling and access to a Smoking Quitline and provided with incentives). The study by Morgan and colleagues148 estimated the effect of a SureStart smoking cessation intervention by estimating the number of women who were reached through the incentive as a proportion of the estimated number of pregnant women smokers living within the SureStart area.

Actors: who the intervention targeted

The majority of studies targeted pregnant women who were smoking (Table 9). Engagement of social support in the intervention occurred in seven studies, although the format of this varied, with five involving ‘social supporters’ or ‘buddies’108,138,141,145,146 and two specifying that the father should perform this role. 104,147 In these studies a social supporter108,138,141,144,146 or partner104,147 was also registered to the programme and in one study141 also received incentives. Peer support was considered an essential element in two of the studies. 104,138 The involvement of the wider community was also a key factor in three of the studies104,108,140 and Cluss and colleagues140 concluded that community-reinforced interventions deliver results that are comparable to those of RCTs. In terms of equity pointers, most of the studies targeted low-income or disadvantaged women (or WIC programme participants); however, in some studies this was unclear. 110,111

Baseline population characteristics

Baseline characteristics data are summarised in Table 9. Several studies reported no baseline details of their study populations,108,142,146,148 although the studies by Lowe and colleagues146 and Walsh and colleagues108 indicated that there were no significant differences at baseline between the control and the experimental groups, without reporting further details.

Of those studies that reported baseline characteristics, with the exception of the study by Ussher and colleagues,150 the mean age of participants was ≤ 30 years. In terms of the ethnicity of participants, over half of the participants in each study population were white, apart from in the studies by Albrecht and colleagues,138 Cinciripini and colleagues139 and Lillington and colleagues,145 which was a study specifically designed to evaluate the effectiveness of a smoking cessation intervention among ethnic minority pregnant smokers.

Twelve studies reported the domestic status of participants,104,105,110,111,138–141,144,147,149,150 with more than half of study participants reported as single (or not married or living with a partner) in eight of the studies. 105,110,138–140,144,149,150 Gestational age was frequently prespecified in the inclusion criteria and therefore the maximum range for this characteristic can be derived from this information even when baseline characteristic data have not been reported, as was the case in the study by Walsh and colleagues,108 although range data were also provided separately too. 138,150 With the exception of the study by Heil and colleagues,105 in which the mean gestational age was 9.2 weeks, the mean gestational age for all studies reporting this information was between 10 and 20 weeks.

For smoking behaviour, the average number of cigarettes smoked per day before discovering they were pregnant exceeded 10 for all studies reporting this information. 105,111,139,143,144,149,150,152 Conversely, the reported average cigarette intake per day since discovering pregnancy was < 10 for all studies104,105,138,139,144,147,150 except for that by Higgins and colleagues,144 although it is worth noting that this study population had the highest mean pre-pregnancy cigarette intake of all those reporting this characteristic.

Apart from these characteristics there was less consistency in reported baseline data across studies; other characteristics reported included income, education, employment and mean Fagerström Test for Nicotine Dependence (FTND) scores. 164 Most studies targeted either disadvantaged areas or disadvantaged women. One study specifically aimed to recruit African American/Hispanic women145 and one was designed for pregnant adolescents who were still at school. 138

Effectiveness

The four category A studies, which compared incentives contingent on validated smoking cessation outcomes with non-contingent incentives for participation in a smoking cessation programme and providing outcome data, were considered sufficiently homogeneous for smoking cessation rates to be combined in a random-effects meta-analysis. 105,109,141,144 Figure 9 summarises the meta-analysis for biochemically validated cessation towards the end of pregnancy (from 7 months’ gestation onwards). The estimated RD was 0.23 [95% confidence interval (CI) 0.14 to 0.31], favouring the provision of contingent incentives. The RR of cessation was 2.58 (95% CI 1.63 to 4.07). All control groups in the studies included in this meta-analysis received some form of non-contingent incentive for participation in a smoking cessation programme and providing outcome data although the value provided varied.

FIGURE 9.

Meta-analysis for the effect of incentives contingent on biochemically validated smoking cessation compared with non-contingent incentives towards the end of pregnancy.

The study by Donatelle and colleagues141 also reported the proportion of those maintaining abstinence at 2 months post partum, with 21% (22/103) of participants receiving partner incentivisation as well as incentives for themselves maintaining abstinence at 2 months compared with 6% (6/102) of participants receiving incentives for themselves alone (RR 3.63, 95% CI 1.54 to 8.85; p = 0.001). At 3 months post partum, Heil and colleagues105 reported that 24% (9/37) of those receiving contingent vouchers were abstinent compared with 3% (1/40) of those receiving non-contingent vouchers (RR 9.72, 95% CI 1.29 to 73.12; p = 0.004). At the same time point, Higgins and colleagues144 reported that 33% (10/30) of participants receiving contingent vouchers were abstinent compared with no participants receiving non-contingent vouchers (p = 0.002).

At 6 months post partum, Heil and colleagues105 reported that 8% (3/37) of those receiving contingent vouchers were abstinent compared with 3% (1/40) of those receiving non-contingent vouchers (RR 3.32, 95% CI 0.36 to 30.58; p = 0.26). Higgins and colleagues144 reported that 27% (8/30) of participants receiving contingent vouchers were abstinent for the same period compared with no participants receiving non-contingent vouchers (p = 0.007).

All category B–D studies reported cessation rates at various time points and are summarised narratively. None of the category B studies comparing incentive interventions (intervention) with no incentive (control) were considered suitable by the research team to be included in the meta-analysis. For example, if other components of the intervention (e.g. counselling) had been provided multiple times it was not suitable to combine these data with data from studies in which the same component had been provided once as the effect of providing the incentive could be confounded with the effect of providing more intensive counselling. Category C studies had no control groups and in category D studies the incentive had been provided to all participants and these studies were therefore unsuitable for meta-analysis.

Smoking cessation outcomes

The effect of incentives on smoking cessation was our main outcome of interest. Apart from the meta-analysed data (see previous section), the reported numbers and percentages of those who quit smoking are provided. This outcome was reported for all studies as this was a prerequisite to meet the inclusion criteria for the effectiveness review. Details of smoking cessation rates at various time points during and beyond pregnancy are provided in Tables 10–15. Outcomes were sometimes reported at time points after the incentive programme had ceased (see Table 8). Intention-to-treat (ITT) and attrition data were often unclear, as detailed in Figure 9. RRs have been calculated from the reported smoking cessation rates and care was taken to use the most conservative estimate of the sample size (e.g. number randomised, number at the start of the study) to prevent any unintentional overinflation of the effectiveness of interventions. However, this was not possible in all cases because of the quality of reporting.

In addition, Lowe and colleagues reported data for a historical control group for their study, published previously,165 which suggested a relapse rate of 35%. Lillington and colleagues145 provided an adjusted relapse rate to account for deception rates among participants, based on a method cited elsewhere. 145,166

McBride and colleagues147 provided data on abstinence among partners. Those whose pregnant partners had received care without partner involvement in the intervention and those who were involved in their partner’s intervention were incentivised groups. Of the 84 men in the partner-assisted group, 15 (17.9%) were abstinent when their partner was at 28 weeks’ gestation and 17 (20.2%) and 20 (23.8%) were abstinent when their partner had reached 2 months and 6 months post partum respectively.

Smoking reduction

Albrecht and colleagues138 reported that those randomised to one intervention group (intervention including social support) smoked on average four fewer cigarettes per day than those in the other intervention group (without social support) and the control group combined. Using digitised data reported in a figure, results indicate that the intervention group smoked an average of 2.8 cigarettes per day post intervention compared with an average of 5.8 in the other groups (i.e. three fewer cigarettes per day on average). Gulliver and colleagues104 reported that respondents who did not quit smoked three to eight cigarettes per day. Pbert and colleagues149 graphically showed the number of cigarettes smoked by those who did not quit. Digitised data for the reduction in cigarette intake are reported in Table 16.

Cluss and colleagues140 reported the number of participants who reduced their cigarette intake even though they did not quit (217/797, 27.3%) but did not provide data on the average number of cigarettes smoked per day for these participants.

Alternative definitions of abstinence

Point-prevalent abstinence was the most commonly used method to define abstinence, likely because of the ability of biochemical measures to validate abstinence only across short time periods. However, several studies defined abstinence in alternative ways to account for consecutive periods of point-prevalent abstinence among participants.

The study by Cinciripini and colleagues139 distinguished between continuous abstinence, defined as no smoking on any day, and prolonged abstinence, defined as not having a relapse of either 7 consecutive days of smoking or smoking at least one cigarette over 2 consecutive weeks within the period of interest. In this study both treatment groups received incentives (as it compared different methods of providing behavioural support).

Continuous abstinence rates ranged from 21.0% (27/129) to 23.4% (30/128) at 3 months following treatment, which was on average 9.5 weeks (SD 4.6 weeks) post partum, from 10.9% (14/129) to 11.7% (15/128) at 3 months post partum, from 8.5% (11/129) to 11.1% (14/128) at 6 months post treatment, which was on average 17.5 weeks (SD 7.3 weeks) post partum and from 1.2% (3/129) to 3.1% (4/128) at 6 months post partum.

Prolonged abstinence rates ranged from 27.1% (35/129) to 31.3% (40/128) at 3 months following treatment (see above for definition), from 16.4% (21/128) to 18.6% (24/129) at 3 months post partum, from 14.1% (18/128) to 14.7% (19/129) at 6 months following treatment (see above for definition) and from 6.2% (8/129) to 7.8% (10/128) at 6 months post partum.

The study by Heil and colleagues105 reported that 24% of women receiving vouchers contingent on smoking abstinence sustained abstinence throughout the third trimester compared with 3% of those receiving non-contingent incentives. However, the time point at which this was established (and therefore the total numbers of women involved) is unclear.

McBride and colleagues147 reported sustained abstinence (‘not smoking at all four follow-ups’) among women but this was measured at 12 months’ follow-up and is therefore outwith the scope of this study. However, the authors also reported the proportion of women who were abstinent at 28 weeks’ gestation who remained abstinent at 2 and 6 months post partum. In total, 55% and 47% of women in the control group sustained abstinence at these time points, respectively, compared with 54% and 52% of women receiving the intervention without partner support and 60% and 52% of women who received both the intervention and social support. However, the total numbers of women in each group at each time point is unclear.

Pbert and colleagues149 reported subgroup data for those women who had spontaneously quit at baseline (i.e. who had been recent smokers until they discovered their pregnancy) who remained non-smokers at 9 months’ gestation and 1, 3 and 6 months post partum. In this study the control group received no intervention whereas the intervention group received vouchers as an incentive for completing interviews regarding their smoking status. Using ITT analysis based on the number of spontaneous quitters at baseline, in the intervention and control groups, respectively, 70.3% (57/81) and 78.2% (60/77) of women remained non-smokers at 9 months’ gestation compared with 40.1% (32/81) and 49.7% (38/77) at 1 month post partum, 12.8% (10/81) and 27.7% (21/77) at 3 months post partum and 11.7% (9/81) and 24.9% (19/77) at 6 months post partum.

Walsh and colleagues108 reported both self-reported and biochemically validated (urine cotinine of ≤ 500 nmol/l) continuous abstinence rates (abstinent on two or three consecutive follow-up visits). In this study the intervention group received incentives whereas the control group did not. Self-reported consecutive abstinence was 13% (17/127; 95% CI 8% to 21%) for participants in the experimental group compared with 3% (4/125; 95% CI 1% to 8%), for participants in the control group from first follow-up to the end of pregnancy, and 8% (10/127; 95% CI 4% to 14%) compared with 1% (1/125; 95% CI 0% to 4%) for the experimental and control groups, respectively, from first follow-up to the post partum follow-up. However, when biochemical validation was used, no control group participants were found to be abstinent on consecutive occasions. For the experimental group, consecutive abstinence from first follow-up to the end of pregnancy was 9% (12/127; 95% CI 5% to 16%) and from first follow-up to the post partum follow-up was 6% (8/127; 95% CI 3% to 12%).

The study by Gadomski and colleagues143 graphically reported time to first failed post partum test for those receiving incentives. Insufficient information was available to estimate data from this graph using digitising software. The study by Heil and colleagues105 reported that, for participants receiving contingent voucher incentives, the mean number of weeks of continuous abstinence as measured at the end of pregnancy was 9.7 weeks (SD 1.9 weeks) whereas, for those receiving non-contingent vouchers, the mean number of weeks of abstinence was 2.0 weeks (SD 0.8 weeks).

Changes in biochemically validated indicators of smoking cessation

Two studies reported changes in biochemical validation readings over the study time frame. The study by Albrecht and colleagues138 reported an average reduction in the CO level of 1.7 p.p.m. among those receiving the intervention (including incentives) plus social support at between 4 and 6 weeks post baseline compared with an increase of 0.9 p.p.m. among those receiving either the intervention (including incentives but without social support) or the control treatment (no incentive or intervention). Mean exhaled CO levels for the groups were also reported graphically and these data confirm the reported rates. The study by Heil and colleagues105 reported the mean decrease in urine cotinine levels from baseline for those receiving contingent voucher incentives (39 ng/ml) and those receiving non-contingent vouchers (33 ng/ml) as measured at the end of pregnancy.

Typology of smoking cessation in pregnancy quitters and non-quitters

Qualitative data from four studies109,110,111,113 illuminate how pregnant smokers are a heterogeneous group who attach different values to incentives and other BCT components. A typology of women engaging in an incentive intervention in Scotland describes six user groups. 113 The least successful at quitting were characterised as ‘breadline survivors’113 who felt pressured to quit109 and who were typically the most socially and financially disadvantaged,109,111 with an unsupportive or no partner. 110,111 They were ‘shifters’ whose changes in smoking behaviour included erratic reductions and increases. 110 Those who were most successful were ‘mothers to be’,113 ‘quitters’,110 with ‘a moral identity’110,111 for whom the child’s health was at the forefront of their decision-making, and they were women who had more stable lifestyles with more control over their environment. 110,113 Some women in a number of studies also reported concern for their baby, concern for self and concern for existing children as reasons for wanting to quit. 109,111 Also successful in the Scottish study were ‘enthusiastic amateurs’,113 who had a history of quit attempts and who benefited most from the support of those delivering the programme and the CO monitoring. 109 These women tended to have more stable lifestyles and to conceptualise the incentive as part of a wider rewards and social support structure,109 particularly valuing relationships established with their local community pharmacists. Women who cut down, ‘harm reducers’,110 or those who relapsed were more likely to have stressful, chaotic lives and perceive smoking as an essential strategy for coping. 110,111

Other groups identified by Radley and colleagues113 were ‘novice quitters’, ‘opportunists’ and ‘impulse shoppers’. ‘Novice quitters’ were typically younger, first-time mothers still living within the family home. They had limited motivation, were looking for a quick solution and were least likely to benefit from the incentive scheme. ‘Opportunists’ saw the scheme as a financial opportunity, had minimal engagement with other intervention components, were often light smokers and were relatively confident of their ability to give up. ‘Impulse shoppers’ were women with limited commitment and those who did not expect to succeed. They had tried to quit smoking numerous times but thought that they should try again as they were pregnant. They were not necessarily looking to participate in a scheme but were speculatively looking for advice. 113

Women seemed not to be aware of the effect of incentives on their motivation, except in one study109 in which some women reported a connection, for example ‘And then the vouchers give me incentive to, like, stop smoking’ (participant 14, incentivised group).

In most studies reporting qualitative data the incentives were described as an ‘added bonus’ in addition to the general BCTs received. 109 Where incentives seemed to play their greatest role, however, was in engaging more women in the study or programme and in achieving higher quit rates. 113 In one study, women in the intervention group who received incentives engaged more with stop smoking services and were more positive about their experiences than those in the control group. 109

Attrition

Attrition rates were reported (or calculable) for 14 studies. Details for different time points ranging from early follow-up in pregnancy through to around 6 months post partum are reported in Tables 17–20.

Higgins and colleagues144 reported compliance as the proportion of monitoring sessions attended and found that 531 of 833 scheduled sessions (63.7%) were attended by women receiving vouchers contingent on cessation and 426 of 673 scheduled sessions (63.3%) were attended by women receiving non-contingent vouchers.

The study by Cinciripini and colleagues139 reported the proportions of respondents completing various numbers of therapy sessions as specified by the intervention, with 78% (208/266) completing seven sessions, 74% (197/266) completing eight sessions and 70% (186/266) completing nine sessions. The authors also reported that there were no between-group differences in treatment session attendance. In any case, in this study both groups received incentives.

In the study by Lillington and colleagues145 incentives were provided for the completion of behaviour change sheets. In total, 88% (136/155) of women completed one or more behaviour change sheet, 32% (50/155) completed between two and seven sheets, 19% (30/155) completed more than seven sheets and 8% (12/155) completed all 12 sheets.

Heil and colleagues105 reported compliance to be relatively high (83–95%) and not significantly different between the two treatment conditions at any assessment point, although no further details were provided.

Lowe and colleagues146 reported details of the proportions of participants who received individual components of the intervention. Of 45 participants, 93.9% (n = 42) received counselling, 97.9% (n = 44) received the calendar, 93.8% (n = 42) received a letter to give to their social supporter/’buddy’, all received a non-smoking contact, 66.7% (n = 30) received a physician’s letter to their home, 68.4% (n = 31) received reinforcement from a doctor, 71.1% (n = 32) received reinforcement from a registered nurse, 86.2% (n = 39) received ‘quit tips’ and 78.8% (n = 35) received deep-breathing instructions.

The study by Lowe and colleagues146 also reported the proportions of participants receiving various kinds of social support from their nominated social supporter/’buddy’. For the intervention group (who received incentives), 72.7% (n = 33) received support from a friend and the same proportion reported that they received help and praise from others and that their significant other understood stress. A total of 78.8% (n = 35) said that they received support from their significant other and 57.6% (n = 26) felt that they had the ability to deal with staying quit. Among the control (non-incentivised) group, 62.2% (n = 32) said that they received support from a friend, 72.4% (n = 38) received help and praise from others and 69.0% (n = 36) received support from their significant other. A total of 58.6% (n = 30) felt that their significant other understood their stress and 44.8% (n = 23) felt that they had the ability to deal with staying quit.

‘Deception’ rates

Some studies have tried to examine the issue of whether women ‘cheated’ (stated that they were abstinent when in fact they had smoked) to receive an incentive. ‘Deception’ was commonly measured by either comparing the difference between self-reported abstinence and biochemically validated abstinence or comparing different types of biochemical validation.

For example, the study by Gadomski and colleagues143 verified negative CO tests by comparing 10% of these samples with a saliva cotinine test to derive a deception rate (the number of positive saliva cotinine tests divided by the number of negative CO tests). For intervention groups there were large differences between samples (from 6% to 61%) but the time point at which these samples were verified was not reported, which is problematic as incentives (vouchers for nappies) were provided to intervention participants only post partum and not during pregnancy.

Walsh and colleagues108 reported inconsistencies between self-reported abstinence and biochemical validation (defined as a urine cotinine level of ≤ 500 nmol/l). Deception rates in self-reported abstinence were 12% for the experimental group (receiving the intervention) and 52% for the control group (not receiving the intervention), although there was some disparity in the proportion of respondents whose self-reported abstinence could be verified (86% in the experimental group and 78% in the control group). Pbert and colleagues149 reported the proportions of participants for whom saliva cotinine values were available, which was 58% at 9 months’ gestation, 49% at 1 month post partum, 43% at 3 months post partum and 46% at 6 months post partum, although it is not clear whether these percentages include those lost to follow-up at each time point. Problems with storage and transfer of samples were reported as the reason for this. Agreement between the self-reported and the cotinine-validated (with ≤ 20 ng/ml classified as abstinence) abstinence rates for the intervention and control groups, respectively, was high except at the final testing point: 89% and 92% at baseline, 84% and 94% at 9 months’ gestation, 81% and 88% at 1 month post partum, 71% and 90% at 3 months post partum and 40% and 80% at 6 months post partum. Both groups in this study received non-contingent incentives.

Lillington and colleagues145 also reported agreement between self-reported abstinence and saliva cotinine-verified abstinence (defined as < 20 ng/ml) at 6 months post partum. However, of the 111 saliva samples available for this time point, 23% (n calculated as 26 participants) exceeded the cut-off (of which 26% were from the intervention group and 21% were from the control group, although the denominators for both groups are unclear). The intervention group received incentives in this study although these were contingent on completing behaviour change worksheets and not on abstinence success.

This study also reported the reasons for not obtaining cotinine samples for validation, of which participant refusal was the reason in 48/143 (33.6%) cases. The study also reported that the intervention group (receiving incentives contingent on completing behaviour change sheets) had a lower refusal rate (23%) than the group not receiving incentives (33%), although the number of participants involved is not clear (numerators and denominators were not reported). 145

Costs

No studies reporting the cost-effectiveness of incentive interventions were identified; however, several other potentially relevant outcomes were reported by individual studies. Cost data were reported by five studies. 104,105,108,140,144 For the studies by Cluss and colleagues140 and Walsh and colleagues108 these data were estimates of the cost of providing the intervention.

For Cluss and colleagues,140 the cost per participant enrolled in the intervention was US$616.91 and the cost per abstainer was US$4433.87. The total cost of the intervention over the course of its 8-year implementation was US$527,631.00. A 2011 price year is assumed. This study had no control group.

For Walsh and colleagues,108 costs were reported as A$13.95 per participant for the intervention group (who received incentives) and A$1.83 for the control group, with the total cost per biochemically validated abstainer as measured at the end of pregnancy of A$121.41 for intervention participants and A$37.88 for control group participants (1997 price year assumed).

Two other studies105,144 reporting information on costs provided data on mean voucher earnings for those receiving vouchers contingent on cessation and for those receiving non-contingent vouchers. For the study by Heil and colleagues,105 voucher earnings as measured at the end of the intervention (12 weeks) in the contingent group ranged from US$0.00 to US$1180.00, with mean earnings of US$461.00 (SD US$456.00), and in the non-contingent group ranged from US$75.00 to US$670.00, with mean earnings of US$413.00 (SD US$163.00). A 2008 price year is assumed. For Higgins and colleagues,144 voucher earnings in the contingent group ranged from US$0.00 to US$1135.00, with mean earnings of US$397.00 (SD US$414.00), whereas voucher earnings in the non-contingent group ranged from US$35.00 to US$517.00, with mean earnings of US$313.00 (SD US$142.00). (A 2004 price year is assumed.)

The study by Gulliver and colleagues104 reported that funding did not exceed US$4000.00 (2004 price year) but noted that this figure did not include donated advertising costs and the costs of the incentives.

Implementation and sustainability

There were no data relevant to the sustainability of the incentive interventions in terms of the long-term benefits outweighing the costs, or to unintended effects once the research has finished. In two studies, incentives donated by local businesses were specifically sought to save costs and to make the programme more sustainable to principal funders. 104,108 Several authors commented that implementation in practice would require provider service changes;142 additional contacts over and above usual care to allow for monitoring/contingency;105,144 the enhancement or intensification of usual care contacts;104,108,143,146 a joined-up approach or case management;111 one-to-one specialist care in pregnancy;148 and additional care post partum to help to prevent relapse. 149 All of these suggestions are likely to have considerable resource implications. However, there was some disagreement in that the authors of two studies concluded that delivery could be effective even in brief (existing) usual care consultations in pregnancy. 145,149

Acceptability outcomes

For some studies, authors had conducted a pilot study to determine feasibility, for example the study by Higgins and colleagues144 was a pilot for the study by Heil and colleagues. 105 Other studies used a process evaluation, which included participants’ feedback,142,145,146,148 or planned to conduct an evaluation. 141 Four studies reported information on the acceptability of the interventions. Walsh and colleagues108 reported that, for midwives, the duration of the intervention was acceptable. This staff group spent significantly more time (t-test value of p < 0.0001 reported) discussing smoking with experimental (incentivised) patients (on average 7 minutes and 12 seconds) than with control (non-incentivised) patients (on average 1 minute 46 seconds).

The study by Cluss and colleagues140 reported that 90% of participants responding to a question about the quality of the service they received were ‘very satisfied’ and 98.5% said that the intervention had helped them deal more effectively with quitting smoking or cutting down their smoking intake. In addition, 95.5% said that they would refer a friend. In terms of the components that respondents found to be the single most helpful treatment, 40.8% cited the interventionist support, 30.5% indicated that CO monitoring helped, 14.8% stated that it was education about the effects of smoking, 5.1% said that it was the improvement to their health, 3.5% cited incentives, 2.1% said that it was problem-solving assistance and 3.3% said that it was another reason.

Ussher and colleagues150 reported that 30 out of 32 recruited participants (94%) had indicated that they would like to join a study involving randomisation to a physical activity (the main component of the intervention in this study) or the control condition.

Lynagh and colleagues151 surveyed acceptability and feasibility of incentives among outpatients at an antenatal clinic of a public hospital in Australia. The self-administered survey was distributed to women between 16 and 42 weeks’ gestation by clinic staff. There were quality limitations with the findings as numerators and denominators were not well reported. It was found that only 25% of 213 participants would endorse the idea of paying women to quit smoking (agreed or strongly agreed). In total, 60% disagreed or strongly disagreed that it was a good idea and 15% were undecided. In terms of feasibility, however, support for the idea that such a programme would be successful was greater and just over half thought that it might work (30%) or were undecided (22%). Among respondents, those who were smokers themselves were significantly more likely to think that incentives were both a good idea (p < 0.0001) and that they would be likely to work (p < 0.003).

Number of studies identified

We identified 5408 records from the primary searches for this review. After title and abstract screening, 5092 studies were considered not to be relevant and were excluded, leaving 316 studies for full-text assessment; in addition, 10 studies were identified from other sources. In total, therefore, the full texts of 326 studies were screened, of which 19 studies (20 reports103,106,107,112,167–182) were included in the review. Two of the studies included in the effectiveness and process reviews were conference abstracts. 180,181 A flow chart of the screening process is provided in Figure 10.

FIGURE 10.

Flow chart outlining the screening process for the review of incentives for breastfeeding.

Included and excluded studies

Nineteen studies evaluating incentives for breastfeeding met our inclusion criteria because an incentive had been provided, even though evaluating the effect of incentivisation was not always the primary purpose of the study. Seventeen full study reports103,106,107,112,167–179 and two conference abstracts180,181 were included in the effectiveness and process review. One additional study was included in the process review only as it did not provide the required breastfeeding outcome data for inclusion. 182 Seventeen studies103,106,107,112,167,169–181 provided incentives to childbearing women and one study168 levied financial penalties against organisations for not meeting their own breastfeeding targets. The study providing incentives to organisations for breastfeeding168 is considered separately later in this chapter.

There were reports in which it was unclear whether the study met our definition of an incentive. Decisions to include as an incentive study were made for the following:

1. Organisational awards/quality standard achievements (e.g. ‘kite marks’) that have a tangible value such as a plaque, certificate or award ceremony, and usually require financial investment to attain, were included and can be considered commitment contracts. An example is the UNICEF Baby Friendly Initiative (BFI),183 which awards hospitals and communities with a plaque and can be considered as a commitment contract. The literature search identified 22 reports describing the BFI, which is a multifaceted and multilevel intervention. Hospitals and primary care organisations pay money to UNICEF UK to complete an accreditation process, commit to training staff and receive an esteemed award on achieving the quality standards. The reports identified by our search strategy represent a small proportion of a substantial body of literature that has recently been systematically reviewed by a team including co-applicant FD. 184 A full review of the BFI literature to identify data relating to the incentive component of interest is beyond the scope of our study because of the large number of publications. The Beake and colleagues184 review is summarised later in this chapter (see Results of incentives provided to organisations to improve smoking cessation in pregnancy and breastfeeding outcomes).

2. As with the review of studies providing incentives for smoking cessation in pregnancy, US WIC programme studies were included if they provided incentives over and above those routinely provided as part of the programme or reported qualitative findings about incentives.

3. Studies providing medical devices to assist the target behaviours that are not routinely provided by the UK health service, which women can keep and which potentially have a monetary exchange value. Studies providing breast pumps were therefore included.

Decisions to exclude as an incentive study were made for the following:

1. Studies providing medical devices to assist the target behaviours that are routinely provided free of charge by the UK health service as a treatment and which women can keep. Nipple shields were therefore excluded.

The reasons for exclusion of assessed full-text papers are given in Figure 10.

Quality of the included studies

Details of the questions used to establish the quality of 16 of the 18 included studies presenting quantitative outcome data are available in Appendix 14. 103,106,107,112,167–176,178,179 The quality of two conference abstracts180,181 was not assessed because of the limited data presented. The study by Wright and colleagues182 was not quality assessed as no appropriate tool was identified to assess the quality of case studies and it provided few data.

Information about quality across studies for each individual aspect of quality and across all aspects of quality for each individual study are provided in Figures 11 and 12 respectively. For both figures, answers of ‘yes’ represented the maximum possible level of quality. Therefore, in considering Figure 11, it can be seen that most studies recruited women at the same point in their pregnancies (88%103,106,107,167–172,174–179), had a follow-up period that was similar between groups (94%103,106,107,112,167,168,170–179) and long enough to detect important effects on outcomes of interest (100%103,106,107,112,167–179), used experienced staff (94%103,106,107,112,167,169–172,174–179), had appropriate facilities to perform the intervention (100%103,106,107,112,167–179) and reported all important outcomes (100%103,106,107,112,167–179).

FIGURE 11.

Quality assessment results by quality appraisal question for all studies included in the review of incentives provided to childbearing women for breastfeeding.

FIGURE 12.

Quality assessment results by study for all aspects of quality for all studies included in the review of incentives provided to childbearing women for breastfeeding.

More problematic aspects of study quality within this group of studies were the lack of description of those lost to follow-up (31% of studies reported this adequately112,167,170,178,179) and an imbalance in demographic characteristics between groups (38% of studies reported this106,107,112,172,174,178); in addition, only two studies conducted analysis based on ITT. 173,174 Only five studies (25%) used a valid and reliable outcome measure. 112,167–169,174 For the purposes of this review, studies were considered to have a valid and reliable outcome measure if they reported a recognised description of how they defined breastfeeding. When studies were assessed for adequacy of reporting on randomisation and recruitment, that is, whether randomisation sequences had been adequately generated for RCTs or whether patients were recruited consecutively for non-randomised studies, only one study was considered adequate. 103 Because of the nature of the interventions, none of the participants was blind to treatment status and in only one study were caregivers blinded. 174

Two independent researchers (VS and SM) appraised the qualitative part of a mixed-methods study in one paper because it had been written by members of the research team involved in quality assessment. 112 The CASP score for this paper was high at 8 out of 10, with half a point given for ethical issues because of a lack of clarity around the consent procedure and only verbal information being provided and half a point given because of weak integration of mixed-methods analysis. No points were given for reflexivity.

Characteristics of the included studies

Initially, studies included in the review of incentives provided to childbearing women for breastfeeding were classified according to study design, as described in the smoking cessation review. This classification is presented in Table 21.

Unlike the incentives in the smoking cessation review, there were no category A studies or other categories of study suitable for meta-analysis. Thus, to maximise the potential of the narrative evidence synthesis, studies were categorised according to the type of incentive, with the order within each category determined, first, by study design, with randomised studies placed before other intervention designs, and, second, by sample size (Table 22). This order is maintained for all tables in this chapter.

The main characteristics of the studies, most of which originated in the USA (17/18), are described in Table 23.

Incentive/reward-related elements

The detail of the incentive interventions provided are summarised in Table 24. Studies provided multicomponent interventions to encourage new mothers to breastfeed, with interventions (or even individual components within the intervention) being provided with varying frequency, intensity and duration (Table 25). Three studies were considered to have more than one incentive intervention arm for the outcome of interest (i.e. breastfeeding). The four-arm RCT by Bliss and colleagues167 compared discharge packs containing manual breast pumps alone, manual breast pumps with infant formula or formula alone with a control group receiving neither formula nor a pump. The RCT by Rasmussen and colleagues174 compared two incentive intervention arms, a manual breast pump or the loan of an electric breast pump for 14 days, with a control group receiving no pump. The non-RCT by Bai and colleagues180 compared discharge packs containing a manual breast pump and breastfeeding information, discharge packs with information only and ‘commercial’ discharge packs that contained infant formula. The incentive components varied across the studies and ranged from a breast pump103,172,174,180 to a combination of a breast pump and feeding-related items,167,169,170 a breast pump, vouchers, gifts and a raffle for both parents,176 enhanced food packages in the context of the WIC programme,171,181 gift items or vouchers107,112,175,179,182 or cash. 173,178 Only one study provided non-material incentives as a help with non-nursing tasks. 106 In all studies incentives were provided at consistent and predictable time points. In addition, in the study by Sciacca and colleagues,176 a raffle was provided at 3 months post partum with a wide range of gifts and experiences on offer.

All incentive interventions involved interactions with incentive providers or other professionals either explicitly or implicitly supporting breastfeeding and we refer to this as the general BCT component of the intervention. These were extracted for each study when reported. However, general BCT components are frequently under-reported and no study mentions observations or recordings of interactions during intervention delivery to assess fidelity or content. We could not extract data for the control arms because of poor-quality reporting, particularly of usual care.

The proportions of the intervention that can be attributed to the incentive and to the general BCT component varied between studies and were sometimes difficult to assess. Across studies, women were invited to participate in a range of activities to obtain the incentives, some of which were reported as usual care plus additional components,103,170,171,179,181 an enhanced programme of care106,112,175,176,178,182 or involved the implementation of a new protocol. 107,167,169,172–174,180

General behaviour change techniques

One study was excluded from BCT extraction as the intervention was rolled out to providers at the health system level rather than at the individual level168 and was thus not in line with the target behaviour. Two studies were excluded from BCT extraction as they were reported only in abstract form, which prevents accurate judging of intervention content. 180,181 Moreover, one study178 targeted partners of pregnant women and BCTs were rated only in relation to increasing breastfeeding in women (i.e. support-related BCTs) rather than rating the BCTs used to target partners’ support behaviours (e.g. information provision and instruction techniques).

Applying the 93-point BCT taxonomy described by Michie and colleagues17 to the included studies resulted in an acceptable inter-rater agreement, with a kappa of 0.77, ranging from 0.69 (‘instruction on how to perform the behaviour’) to 0.88 (‘adding objects to the environment’). Explicit reporting of BCTs underpinning interventions varied and reporting was often unclear or implicit. Altogether, 27 study arms were rated for inclusion of 82 BCTs, including seven control/usual care arms and 20 intervention arms. Intervention arms used an average of 4.34 (SD 2.8) BCTs, ranging from one170,173 to 12,175 whereas control/usual care arms used an average of 1.7 (SD 2.5) BCTs, ranging from zero106,107,175 to five. 176

Out of 82 BCTs rated, study arms were identified to have used 22 (Figure 13), with the BCTs used most often used being ‘Adding objects to the environment’ (n = 14), which, in most studies, took the form of providing breast pumps, ‘Social support (unspecified)’ (n = 12), ‘Instruction on how to perform the behaviour’ (n = 11) and ‘Information about health consequences’ (n = 10). As in the smoking cessation review, WIC studies were recorded separately without inferring the presence of specific BCTs given the variation in WIC programme implementation. Six studies were rated as being engaged in the WIC programme171,172,175,176,181,182 in addition to the identified BCTs.

FIGURE 13.

Behaviour change techniques used in studies included in the review of incentives for breastfeeding.

Incentive and reward elements were identified reliably (kappa = 0.73). Out of the 18 studies, two176,178 used multiple incentives within an intervention, which varied across the four assessed incentive elements: contingency target, actor, type and content (see Table 24). Two studies used different rewards/incentives for different experimental groups within the same study. 172,174 The majority of studies were rated as unclear in relation to whether participants were aware of the incentives/rewards (n = 13), with the remaining rewards/incentives rated as being known to participants (n = 5).

General BCT components included peer support or counselling in five studies, although the formality of this varied from its being a key component, that is, having a nominated peer supporter,107,112,176 to the intervention involving membership of a peer support group107,179 or a trained Hmong peer counsellor. 175 Professional support was offered in one study182 and professional counselling was offered in a number of studies. 106,107,169,175 Proactive telephone calls were made in five studies, with the purpose being predominantly to collect outcome data167,170,173,180 although in one study new breastfeeding mothers were offered reactive breastfeeding support by telephone. 182

Setting and provider

The studies included in the review were typically conducted in the USA (17/18), with only one carried out in the UK112 (see Table 23). Several studies were funded by private sources106,107,167,170,172,178,179,182 whereas others were funded by private and university funds180 or national health or government institutes,175,176,185 one of which explicitly reported the receipt of extra financial support from the WIC programme. 175 One study was jointly funded by a charity, breast pump distributors and local insurance companies, relying on donations from local businesses to provide the incentives. 103 Authors in one study declared research consultancy for a breast pump manufacturer169 and another involved the breast pump manufacturer for distribution. 174 In four studies the funding source was unclear. 169,171,173,181 For the study conducted in the UK, funding was received from the NHS and a breastfeeding charity. 112

The lead authors’ disciplinary backgrounds were also assorted, including lactation specialists,107,179 nutritionists,175 paediatricians/maternal health clinicians,169,170,172,178 public health/health research researchers112,182 and psychologists,170 and those with a health education background. 106,171 Five study teams involved members from across these disciplines. 103,167,174,176,180 Details were not reported for two studies. 172,181

The intervention settings and providers delivering the interventions varied across the included studies (see Table 25). Six of the studies recruited women in WIC clinics and provided the intervention as a supplement to usual care. 171,172,175,176,181,182 Similarly, other studies recruited in centres where concurrent contact with pregnant women was already taking place. 103,106,112,167,170,173,179,180 One study was conducted with employees in a work setting169 and one was conducted at a school with adolescent school pupils who were pregnant. 107

Little information was given regarding the process of providing incentives to participants (see Table 25). Some differentiation was made between the staff delivering the incentive and the staff delivering the general BCT components of the intervention, although in most cases this was carried out by usual care health professionals: WIC staff,172,176,181 doctors and midwives,103,106,173 lactation consultants or nutritionists,107,169,179 hospital staff,167,180 trained interviewers or counsellors,112,170,171,178 peer supporters112 and research staff. 174,182 The details reported were often unclear but some noted when staff were specifically trained to deliver the intervention. 170,175

Intervention intensity and the participant journey

Intensity refers to the number of contacts involved for patients from the start of the intervention until the final outcome data are collected. Studies provided multicomponent interventions to encourage women to breastfeed, with interventions (or even individual components within the interventions) being provided at varying frequencies and time points, and this is summarised in Tables 25 and 26.