Analytical approach:
The economic evaluation was based on a cluster-randomised trial conducted between 2007 and 2009 in Finland. The time horizon was from eight to 12 weeks gestation until discharge from hospital after birth. The stated perspective was societal.
Effectiveness data:
The trial included 399 patients from 14 maternity clinics. These clinics were matched into pairs, for randomisation, based on the number of births, socioeconomic status of patients, incidence of gestational diabetes, and size of the area's population. The effectiveness of the intervention was measured by the child's birth weight, in grams of weight avoided. This was from the maternity cards of patients in the trial and Finnish registry data. Other trial outcomes were physical activity, and quality of life on the 15D questionnaire and on the 10cm visual analogue scale (VAS). These questionnaires were completed by the women at eight to 12 weeks and 36 to 37 weeks of gestation.
Monetary benefit and utility valuations:
Patients valued their utility using the 15D health-related quality of life questionnaire. Missing utility data were imputed using multivariate linear regression.
Measure of benefit:
The mean birth weight and perceived health and quality of life were the measures of benefit.
Cost data:
The analysis included the hospital costs of in-patient and out-patient days, delivery, medical salaries, administration and neonatal care; the intervention costs of nurse and physiotherapist time; and productivity lost. All cost items were measured in the trial. Resource use was from nurse and patient reports, and medical registers. The hospital costs were Finnish national averages from the hospital where 91% of infants were delivered. Productivity lost was recorded on questionnaires completed by the patients every trimester. Patient salaries were based on national averages multiplied by 1.3 to include related expenses. All costs were presented in 2009 Euros (EUR).
Analysis of uncertainty:
Bootstrapping was used to estimate the uncertainty around the costs, and confidence intervals around the incremental cost-effectiveness ratios, as well as to conduct probabilistic sensitivity analysis. The methods adjusted for the effects due to cluster randomisation. The results of the probabilistic sensitivity analysis were presented on cost-effectiveness planes and in cost-effectiveness acceptability curves. Sensitivity analysis assessed the effect of doubling the intervention costs. A subgroup of women who complied with the programme was analysed.