Cost-benefit and extended cost-effectiveness analysis of a comprehensive adolescent pregnancy prevention program in Zambia: study protocol for a cluster randomized controlled trial

Economic evaluations inform resource allocation decisions by generating evidence about the optimal investment of scarce resources in one program or another, by comparing their costs and the potential returns for the investment. It is for this reason that economic evaluation is increasingly becoming an important criterion to inform policy and priority decisions within health systems and other sectors. A number of adolescent pregnancy prevention programs have been tested [15, 16]. However, economic evaluations of these programs are scarce [17, 18], despite the fact that they require substantial investment for scale-up. In this study, we will conduct a cost-benefit analysis (CBA) to compare costs and benefits of two adolescent pregnancy prevention package programs in Zambia. CBA, unlike other types of economic evaluation, will enable us to compare benefits across policy sectors and to capture broad program benefits from reduced adolescent birth complications and increased school completion rates.

We will also conduct an extended cost-effectiveness analysis (ECEA) that builds on the standard cost-effectiveness analysis as a quantitative method for evaluating health policies, inter-sectoral policies and policy instruments that impact population health. ECEA takes into consideration the health and financial consequences of policies including the resulting distribution of financial risk protection from catastrophic health expenditure across socio-economic groups [19]. Maternal and childcare-induced catastrophic health expenditures are common in developing countries [20, 21], despite presence of policies that advocate for free services. It is likely that a program that reduces adolescent childbearing could also prevent the catastrophic health expenditures associated with birth complications. In Zambia, surveys have shown that about 10% of households experience catastrophic health expenditures for outpatient visits alone [22]. Catastrophic health expenditure associated with adolescent childbearing are more likely to occur among the poorest households and will push them into a downward spiral of greater impoverishment thereby worsening social inequity.

The trial is being conducted in 12 districts in rural Zambia: Kalomo, Choma, Pemba, Monze, Mazabuka, Chikankata, Chisamba, Chibombo, Kabwe rural, Kapiri Mposhi, Mkushi and Luano. The districts were selected because they have medium school drop-out rates, and adolescent marriage and childbearing are common. Approximately 22% of girls in these districts have given birth by age 17 and 35% by age 18 years (Zambia Census of Population and Housing (ZCPH), 2010).

The trial aims to compare the effectiveness of economic support with or without a corresponding community dialogue component, in reducing adolescent childbearing and in increasing grade 9 completion rates. The interventions were launched in September 2016 and will continue until November 2018, which is the time when the girls in the study are expected to complete grade 9 (Fig. 1). The interventions are being delivered by teachers and community health assistants (CHAs) or community health workers (CHWs). This will make the results realistic and relevant for any potential scale-up in Zambia or other similar contexts.

Data to be used in the economic evaluation will be collected alongside the main trial. Data on the status of school attendance, childbearing, birth complications and utilization of healthcare facilities are being collected by trained female research assistants through face-to-face follow-up interviews every 6 months. Cost data will be collected by the same research assistants after receiving training on the costing methodology and data collection tools. The data collection team is independent of the intervention delivery team, and the research assistants in the final interview round will be unaware of the intervention status of respondents.

CBA is the only type of economic evaluation that measures expected benefits associated with a program in monetary terms. Further, CBA converts all costs and benefits into a common monetary metric that could facilitate the merging of diverse outcomes spanning different sectors. Program outcomes that will be considered for CBA include increased school completion rates and reductions in adolescent childbirth. The increase in school completion rates is expected to increase returns to education while the reduction in adolescent childbearing is expected to save healthcare costs that would otherwise have been incurred in the management of the associated birth complications.

The expected health gains and non-health gains of the program in their natural units of measurement will be evaluated in ECEA. The expected health gains are adolescent childbirths averted, and non-health gains are private/out-of-pocket (OOP) expenditures averted and financial risk protection. ECEA will also examine the disparities in health gains, OOP expenditures and financial risk protection across different socio-economic groups, and aims to assess the distributional impact of the interventions. We will use the household consumption data to rank the study populations into wealth quintiles, from the poorest to the richest.

Consumption rather than income is a preferred method for measuring standards of living in developing countries because a majority of people are employed in the informal sector, prominent production of goods at home and multiple and often constantly changing sources of income [34]. In addition, the RISE program will provide economic support to families receiving the intervention packages, and this will influence their consumption pattern. Therefore, we will conduct a household survey in order to collect consumption data that can be used to construct living standard quintiles. The data can also be used to assess levels of OOP health expenditure relative to total expenditure. Thus, the survey questionnaire will capture the consumption of food items, non-food items and consumer durables.

The category “food items” will cover both homemade food items and those purchased from the market. Non-food items will be categorized into two groups based on the frequency of purchase. Frequently purchased items will include transport, cleaning supplies, air-time vouchers, cigarettes, tobacco etc. Non-frequent non-food items will include utilities (electricity, kerosene, water, sanitation), clothes, cell-phones, health i.e. expenses on consultations and diagnostics, medicines and bed days, education (school fees, uniforms, exercise books etc.), contributions etc. Expenditures on consumer durables such as housing will be estimated using rental charges or annualized replacement costs if the dwelling is owned by the household. For frequently used items we will use a recall period of 2 weeks [34].

Sample size calculation for the trial has been explained in detail elsewhere [23], and was based on three primary outcomes namely: incidence of births before a girl’s 18^th birthday, incidence of births within 8 months after the end of the intervention period, and the proportion of girls who sit for grade 9 examinations. Reduction in incidence of births and increased proportion of girls completing grade 9 are associated with healthcare cost-saving and increased returns to education, which are short and long-term benefits. The trial has a sample of 157 clusters (~4900 girls). This involves 63 in each intervention arm (1950 girls) and 31 in the control (1000 girls). It gives power of >95% to detect a difference of -40% in the “incidence of births before a girl’s 18^th birthday” in the combined vs control arm, power of 80% to detect a difference of -25% in the economic vs control arm and power of 70% to detect a difference of -20% in the interventions arms. For the outcome “proportion of girls who sit for grade 9 examinations”, this sample produced power of >95% to detect an assumed difference of +27% in the combined vs control arm, +15% for economic vs control and +10% between the intervention arms.

For CBA, the net social benefit (NSB) of each intervention package will be calculated by subtracting program costs from the monetary benefits. The intervention package with the best economic return will be recommended as the optimal choice for scale-up [39]. We will also calculate the benefit-cost ratio for each intervention package by dividing the incremental monetary benefits to the incremental costs relative to the control arm. All costs and benefits will be converted to present values using the commonly applied annual discount rate of 3%. Modeling of returns to investment in education will be performed in TreeAge Pro software (TreeAge Software, Inc. Williamstown, PA, USA). The net benefit will be calculated using Multilevel Models to account for clustering in the trial data [40, 41]. Net benefit will be used as an outcome variable and the intervention arm as a predictor variable. The school identification variable will be used to adjust for clustering.

For ECEA, we will use the method proposed by Xu [42], which assumes that the subsistence needs represent food spending adjusted for household size, and this is considered to be the median total household spending. Any household whose out-of-pocket health payments exceeds 40% of its capacity to pay, i.e. the total household spending less subsistence spending will be categorized as having incurred catastrophic health expenditures. ECEA calculations will be performed with the R program. We will plot concentration curves, i.e. a plot of cumulative proportions of population ranked from the poorest to the richest against cumulative proportion of adolescent births. We will also calculate concentration indices to measure levels of inequality between the trial arms.

Sensitivity analyses will be performed by varying key parameters within plausible ranges to test the robustness of the results. Net benefit and cost-benefit ratio will be re-evaluated by varying program implementation costs, healthcare costs, loss in productivity, discount rates, wages and the overall change in the effectiveness of the interventions. The same will be done for the extended cost-effectiveness analysis.

Economic evaluations will be conducted from both the program and societal perspectives. For the program perspective we will only consider costs related to the implementation of the interventions, which could be borne by the government or any other interested organizations. For the societal perspective we will consider all costs, which include program costs, healthcare costs and household costs.