• Heather Morgan, Pat Hoddinott, Gill Thomson, Nicola Crossland, Shelley Farrar, Deokhee Yi, Jenni Hislop, Victoria Hall Moran, Graeme MacLennan, Stephan U Dombrowski, Kieran Rothnie, Fiona Stewart, Linda Bauld, Anne Ludbrook, Fiona Dykes, Falko F Sniehotta, David Tappin, and Marion Campbell.
• Health Services Research Unit, University of Aberdeen, Aberdeen, UK.
• Health Technol Assess. 2015 Apr 1; 19 (30): 1-522, vii-viii.

BackgroundSmoking in pregnancy and/or not breastfeeding have considerable negative health outcomes for mother and baby.AimTo understand incentive mechanisms of action for smoking cessation in pregnancy and breastfeeding, develop a taxonomy and identify promising, acceptable and feasible interventions to inform trial design.DesignEvidence syntheses, primary qualitative survey, and discrete choice experiment (DCE) research using multidisciplinary, mixed methods. Two mother-and-baby groups in disadvantaged areas collaborated throughout.SettingUK.ParticipantsThe qualitative study included 88 pregnant women/recent mothers/partners, 53 service providers, 24 experts/decision-makers and 63 conference attendees. The surveys included 1144 members of the general public and 497 health professionals. The DCE study included 320 women with a history of smoking.Methods(1) Evidence syntheses: incentive effectiveness (including meta-analysis and effect size estimates), delivery processes, barriers to and facilitators of smoking cessation in pregnancy and/or breastfeeding, scoping review of incentives for lifestyle behaviours; (2) qualitative research: grounded theory to understand incentive mechanisms of action and a framework approach for trial design; (3) survey: multivariable ordered logit models; (4) DCE: conditional logit regression and the log-likelihood ratio test.ResultsOut of 1469 smoking cessation and 5408 breastfeeding multicomponent studies identified, 23 smoking cessation and 19 breastfeeding studies were included in the review. Vouchers contingent on biochemically proven smoking cessation in pregnancy were effective, with a relative risk of 2.58 (95% confidence interval 1.63 to 4.07) compared with non-contingent incentives for participation (four studies, 344 participants). Effects continued until 3 months post partum. Inconclusive effects were found for breastfeeding incentives compared with no/smaller incentives (13 studies) but provider commitment contracts for breastfeeding show promise. Intervention intensity is a possible confounder. The acceptability of seven promising incentives was mixed. Women (for vouchers) and those with a lower level of education (except for breastfeeding incentives) were more likely to disagree. Those aged ≤ 44 years and ethnic minority groups were more likely to agree. Agreement was greatest for a free breast pump and least for vouchers for breastfeeding. Universal incentives were preferred to those targeting low-income women. Initial daily text/telephone support, a quitting pal, vouchers for > £20.00 per month and values up to £80.00 increase the likelihood of smoking cessation. Doctors disagreed with provider incentives. A 'ladder' logic model emerged through data synthesis and had face validity with service users. It combined an incentive typology and behaviour change taxonomy. Autonomy and well-being matter. Personal difficulties, emotions, socialising and attitudes of others are challenges to climbing a metaphorical 'ladder' towards smoking cessation and breastfeeding. Incentive interventions provide opportunity 'rungs' to help, including regular skilled flexible support, a pal, setting goals, monitoring and outcome verification. Individually tailored and non-judgemental continuity of care can bolster women's capabilities to succeed. Rigid, prescriptive interventions placing the onus on women to behave 'healthily' risk them feeling pressurised and failing. To avoid 'losing face', women may disengage.LimitationsIncluded studies were heterogeneous and of variable quality, limiting the assessment of incentive effectiveness. No cost-effectiveness data were reported. In surveys, selection bias and confounding are possible. The validity and utility of the ladder logic model requires evaluation with more diverse samples of the target population.ConclusionsIncentives provided with other tailored components show promise but reach is a concern. Formal evaluation is recommended. Collaborative service-user involvement is important.Study RegistrationThis study is registered as PROSPERO CRD42012001980.FundingThe National Institute for Health Research Health Technology Assessment programme.

12 keywords...

hide…