Introduction

Scientific evidence confirms that the first 3 years of life are when risk factors such as inadequate parental stimulation exert the greatest harm on child development and when effective interventions can have the greatest benefit [1,2]. Numerous studies have demonstrated the effectiveness of parenting interventions that promote responsive stimulation to improve early childhood development (ECD) in low- and middle-income country (LMIC) settings among children under 3 years of age [3,4]. However, very few studies include accompanying economic evaluations to provide information about a program’s cost-effectiveness or the potential economic returns to scaling these interventions in resource-limited settings [5,6]. For example, while recent meta-analyses have identified more than 40 studies evaluating early childhood stimulation interventions targeted to young children in LMICs [4,7], another recent meta-analysis of economic evaluations of ECD parenting interventions and programs identified just 5 studies that provided information about program costs [8]. Institutions such as the World Health Organization (WHO), United Nations International Children’s Emergency Fund (UNICEF), and the World Bank now advocate filling this gap via collection and sharing of program cost data from evaluations of ECD interventions in LMIC settings [9].

Data on program costs can enable different types of economic evaluations of ECD interventions. Cost-effectiveness analysis (CEA) compares the relative costs of a program to its generated impacts on outcomes and is one of the most important indicators used to inform scalability [10]. However, existing CEAs of ECD interventions from LMICs are more readily available for nutrition or health interventions or for preschool programs focused on older ages, with less attention paid to responsive parenting interventions for young children [4,8,11]. Moreover, to date, the very few parenting interventions including a CEA adopt a provider’s costing perspective, thereby excluding any societal costs such as opportunity costs to participants and delivery agents. This risks creating perverse incentives for governments to offload such costs onto users and goes against recommended best practice [12]. Notably, no CEA studies are available from sub-Saharan Africa (SSA), where the coverage of ECD programs for young children is lowest, and supporting evidence to expand coverage of responsive parenting interventions is arguably most needed [10].

A more comprehensive economic evaluation may also include a benefit–cost analysis (BCA), which compares the monetized expected program benefits relative to its costs, and can better inform policy on the potential benefits to society of scaling a program. Yet BCAs are even more rare than CEAs in the evaluation of ECD parenting programs in LMICs [13], likely because it can be difficult to estimate long-term benefits without making a number of assumptions (or waiting many years to measure those benefits). A recent systematic review identified only 3 evaluations of ECD parenting programs focused on young children from LMICs that provided benefit–cost ratios (BCRs) [8]. Again, none are from SSA.

In addition to a lack of economic evaluations of ECD parenting programs, there has been growing interest in understanding the delivery format that is most scalable to reach young children before age 3 in LMICs [14,15]. ECD parenting programs that have been shown to be effective are typically delivered in one of 2 formats: individual home visits or group sessions held in the village or at local health clinics. Home visits have the advantage of offering personalized coaching, demonstration, and feedback, but are time and labor intensive and hence expensive to deliver at scale. Group sessions are less labor intensive and may be relatively less expensive to deliver, can encourage peer support, and can help modify group norms for child rearing, but may be weak to overcome personal barriers to behavioral change [3]. Knowing the cost-effectiveness of competing delivery models can help policymakers select the model that is most scalable for the local context.

The aims of this study are to perform an economic evaluation of an effective ECD parenting intervention targeted to families with children under age 3 in SSA that can contribute new insights about the relative cost-effectiveness of 2 different delivery models and to estimate their potential BCRs and returns to investment.

Materials and methods

Background of research trial

The intervention, named Msingi Bora (“Good Foundation” in Swahili), was a group-based parenting intervention implemented among rural Kenyan households with children aged 6 to 24 months at baseline that significantly improved child cognitive, receptive language, and socioemotional development among treated families [15]. The cluster randomized controlled effectiveness trial of Msingi Bora was conducted across 60 rural villages in western Kenya in the subcounties of East and South Rachuonyo within Homa Bay county and Sabatia subcounty within Vihiga county. The trial featured 3 study arms, each 20 villages in size: (1) a group-only delivery model with 16 fortnightly group sessions; (2) a mixed-delivery model combining 12 group sessions with 4 home visits; and (3) a control group that received no program under the status quo. Eligible participants within villages were mothers or other female primary caregivers aged 15 and over with a child between 6 and 24 months at recruitment without signs of severe mental or physical impairment. All participants provided written informed consent at the time of data collection. At baseline (October 1 to November 12, 2018), we collected household sociodemographic information and measures of child developmental outcomes from 1,152 eligible households. The 60 villages were then randomly assigned to one of the 3 study arms. The intervention lasted from mid-November 2018 to mid-July 2019 and was immediately followed by an endline survey (August 5 to October 31, 2019) among 1,070 retained households. The primary endpoints were child cognitive and language development as measured using the Bayley Scales of Infant Development, Third Edition [16] and child socioemotional development measured with the Wolke Scale [17]. Ethics approval was obtained from the ethics committee at Maseno University in Kisumu, Kenya and RAND; the trial is registered with ClinicalTrials.gov NCT03548558, and the study protocol has been published [18].

Delivery agents were drawn from the existing cadre of local community health volunteers (CHVs), which is a part-time and voluntary position under Kenya’s Ministry of Health, tasked with improving community health through home visits. For our study, CHVs received a total of 16 days of training, with 8 days before the program started to cover the first half of sessions (sessions 1 to 8) and another 8 days at midline to cover the second half of sessions (sessions 9 to 16). Each training included 5 days in the city of Kisumu to introduce the material and curriculum, followed by 3 days of supervised practice in the rural subcounties. From session 4 onwards, CHVs received monthly 1-day refresher trainings in their subcounties to prepare them for that month’s upcoming sessions.

Mothers and children were invited to attend every session and received a small gift for attendance (e.g., small bar of soap or bag of milk). Every fourth session served as a review session, for which households in the group-only arm continued with group meetings, while households in the mixed arm received individual home visits. During these home visits, participants received identical messages to those in the group reviews, but the focus was personalized on that family. The remaining 12 non-review sessions were delivered in group meetings in both intervention arms. Group sessions took place in local community centers or churches and lasted a median of 90 minutes, while the median home visit lasted an hour. The median travel time for CHVs to group meetings was 20 minutes each way, while the median travel time to each home visit was also 20 minutes. Over the 8-month program, mothers in the group-only arm averaged 64% attendance versus 74% in the mixed-delivery arm (Table 1), where this difference was driven only by higher attendance (88%) to the 4 review sessions that were delivered as home visits in the mixed-delivery arm (and thus, mothers did not have to travel to these sessions) [19].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Background information on the intervention arms and control group.

https://doi.org/10.1371/journal.pmed.1003746.t001

The local nongovernmental organization (NGO) Safe Water and AIDS Project (SWAP) oversaw the study’s implementation. Three of SWAP’s staff from the central office in Kisumu provided oversight of the project and became lead trainers of CHVs under a Training of Trainers (ToT) model. They also coordinated the work of 3 subcounty supervisors who were hired by the research project to observe and monitor each group session and at least 1 home visit from each review session in the assigned villages. Subcounty supervisors were assisted by 1 or 2 “mentor” CHVs who had been previously trained in a piloting phase to deliver the program across 6 villages not included in the main trial. In the group-only arm, 89% of the non-review and 91% of the review sessions were supervised. In the mixed arm, 87% of the group non-review sessions and 8% of the 1,531 attempted home visits were supervised (Table 1). Supervisors rated CHVs using monitoring forms with a checklist of items related to fidelity and quality of delivery. CHVs were provided supervisor feedback immediately after each session and filled out self-evaluation forms after each session.

Costs

Provider costs relevant to implementing the Msingi Bora program were collected during the 8-month intervention period using quantitative financial data reported by SWAP. For these costs, we used a step-down accounting costing methodology using actual program costs rather than budgeted costs [24,25] and where costs were divided into direct and indirect implementation costs depending on whether they were related to specific implementation activities or not. All program-related cost data were monitored on a regular basis during the start-up and intervention period by SWAP’s financial manager. We used economic costing methods to estimate opportunity costs for CHV and mother participants. Travel times to conduct sessions and session duration both for home visits and group meetings were based on CHV self-assessment forms collected after every session. We used supervisor monitoring forms to impute supervision time-related costs.

Unit costs by intervention arm are presented in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Unit costs of Msingi Bora.

https://doi.org/10.1371/journal.pmed.1003746.t002

Provider costs.

We present the costs that were required to implement the intervention “as it happened,” which include additional costs incurred due to unexpected delays in fieldwork, as well as cost inefficiencies in practice (e.g., having more than 1 supervisor at a given session). The only exception is that we account for the higher time use costs for CHVs delivering home visits for the 4 review sessions in the mixed-delivery arm using estimates of the opportunity cost of their time. For the sake of comparison, in S2 Text, we present an alternative “best case” costing scenario based on our calculations of program costs without those delays and inefficiencies that better reflect how the program was intended to be implemented and thus how much it would arguably cost to replicate it in similar settings taking advantage of potential economies of scale. Results from this comparison are included in a sensitivity analysis.

Direct program costs.

Direct costs included salaries and stipends for personnel (CHVs, mentor CHVs, subcounty supervisors, and SWAP staff), training, travel and accommodation for training, and any food and supplements necessary for the implementation of the program.

Personnel

We include the flat monthly stipends paid to CHVs and mentor CHVs, as well as salary costs for 3 subcounty supervisors and 4 SWAP staff members involved in day-to-day activities of implementation, supervision, and training, prorated to their percentage time commitment to the project. These base salary costs were equally shared among the 2 intervention arms. For the 4 review sessions that were delivered through home visits in the mixed-delivery arm, we add to these salary costs the time use costs that CHVs, mentor CHVs, and subcounty supervisors spent delivering or supervising these sessions using an estimated hourly wage rate and the time taken to conduct these activities (Table 1). These time use costs were larger in the mixed-delivery arm because home visits are more time intensive than group meetings. To estimate the opportunity costs of CHV’s time, we use the 2015 to 2016 Kenyan Integrated Household Budget Survey (KIHBS), which provides individual data on wages at the county level. We calculate the average implied hourly wage for all workers who work at least 20 hours per week in some outside paid activity in the study counties of Homa Bay and Vihiga, updated to 2020 prices. We used session monitoring data to impute supervision costs for mentors and supervisors as appropriate.

Travel and accommodation

These costs include transportation, meals, and lodging for CHVs, mentor CHVs, and subcounty supervisors to attend the 2 centralized 5-day training sessions in Kisumu, transportation costs for supervisors to supervise sessions, and transportation and meal costs for mentor CHVs and CHVs to attend the monthly refresher trainings hosted in the respective subcounties. These costs again differ by delivery arm due to the travel time necessary to conduct home visit review sessions in the mixed-delivery arm, again priced at local wage rates.

Supplements and incentives

We include milk, soap, or eggs given to participants during sessions, as well as a locally published picture book given to all children in session 14.

Indirect program costs.

These costs included start-up costs such as the cost of tablets used for session monitoring and an online data transfer platform, administrative costs such as the design and printing of manuals for CHVs used during training, mobile phone credit for CHVs, mentor CHVs, and supervisors, and the printing of T-shirts for CHVs and certificates for participants who completed the program.

Societal costs.

In addition to the provider costs above, we include the following societal costs.

Mothers’ opportunity cost of time

Mothers who participate in Msingi Bora pay opportunity costs of their time stemming from attending sessions, as well as enacting the new recommended behaviors with their children at home. To price this time, we again use the 2015 to 2016 KIHBS to estimate an implied hourly wage for all mothers based on the average wage rate for females in the 2 study counties, regardless of their employment status, and updated to 2020 prices (Table 2). We combine average attendance, session duration, and travel times from Table 1 to estimate the time mothers spent attending sessions, and then in response to a reviewer recommendation, we add to this an estimated 1 additional hour per week in stimulation practices over 3 years of the child’s life to reflect time opportunity costs associated with maternal behavioral changes. This implicitly assumes no opportunity costs for behavioral changes after 3 years, but aligns with Kenya’s 2018 policy of universal free preschool beginning at age 3 [26]. Moreover, our curriculum emphasized that mothers can stimulate their children while doing their daily chores such as speaking to the child and providing the child with a playing object and did not ask for more than 10 minutes per day of direct stimulation activities.

Venues

While our program enjoyed the free use of community centers or churches for the group-based sessions, under a scaled version of Msingi Bora, we estimate that venues would no longer be free and thus include a daily rental fee per venue per group session, for which there are 12 in the mixed-delivery arm and 16 in the group-only arm (Table 2).

Results

Program costs

The total costs for the program were US$56,171 for the group-only arm and US$58,103 for the mixed-delivery arm (Table 3), which translate into a total cost per child of US$140 in the group-only arm and US$145 in the mixed-delivery arm. Provider costs comprised the bulk of these total costs, amounting to US$47,423 in the group-only arm (US$119 per child) and US$48,670 in the mixed-delivery arm (US$122 per child). The only difference in these costs across the 2 models of delivery comes from the time use costs associated with the review sessions that account for CHVs spending an average of 18.5 additional hours delivering home visit review sessions relative to group review sessions. Travel and accommodation were by far the highest direct provider costs, largely as a result of the 2 centralized full board trainings. The small difference in these costs across the 2 delivery models stem from the extra travel time required for home visits, which we estimate at an additional 8.5 hours per CHV per review session in the mixed-delivery arm.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Total costs of Msingi Bora.

https://doi.org/10.1371/journal.pmed.1003746.t003

The present period societal costs were US$8,748 in the group-only arm (US$22 per child) and US$9,433 in the mixed-delivery arm (US$24 per child) (Table 3). This small difference stems from a few countervailing factors. On the one hand, the group-only arm entails 4 extra group review sessions that require greater venue costs, as well as greater time for mothers to travel and attend group sessions, which lasted 30 minutes longer than home visits on average. On the other hand, only mothers who attend sessions can enact the recommended behavioral changes, which implies greater mother opportunity costs on average due to the higher attendance at the 4 home visit review sessions in this arm.

Cost-effectiveness

Table 4 reports the ICERs for each delivery model and considers costs under both provider and societal perspectives as reported in Table 3. Higher ratios imply a higher impact on child outcomes per US$100 invested. ICER estimates for the group-only arm are generally greater than those for the mixed-delivery arm due to the higher estimated impacts on child outcomes at slightly lower costs. For the group-only delivery model under a societal perspective, an additional US$100 investment is estimated to lead to a 0.37 SD improvement in child cognition, 0.30 SD improvement in receptive language, and 0.16 SD improvement in socioemotional development. Under the mixed-delivery model, the corresponding estimates are improvements of 0.23 SD, 0.14 SD, and 0.15 SD, respectively. ICERs under a provider perspective are slightly larger as societal costs are excluded from the denominator.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Cost-effectiveness of Msingi Bora.

https://doi.org/10.1371/journal.pmed.1003746.t004

BCRs and ROIs

We estimate gains in lifetime wages from participation in the program of US$2,729 per child in the group-only arm and US$1,784 in the mixed-delivery arm (Table 5), a difference primarily due to the lower intervention impacts on cognition in the mixed-delivery arm. We estimate future schooling costs of US$35 in the group-only arm and US$23 in the mixed-delivery arm, implying total long-term costs per child of US$176 and US$168, respectively. As a result, Table 5 shows that the benefits of Msingi Bora are estimated to substantially outweigh the associated costs under both delivery models, with a BCR of 15.5 and an ROI of 127% for the group-only delivery model and a corresponding BCR of 10.6 and an ROI of 55% for the mixed-delivery model.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. BCRs and ROI of Msingi Bora.

https://doi.org/10.1371/journal.pmed.1003746.t005

Sensitivity analysis

Table 6 shows the calculated values of the key input parameters needed to reach an estimated BCR less than 1 or an estimated ROI less than 0, holding all other values constant. These “extreme values” identify at which point scaling Msingi Bora would become a poor societal investment. For example, to obtain a BCR less than 1 for the group-only delivery model, we would need a discount rate of 17% (as opposed to 5%), holding all other parameters constant at assumed values as listed in column 1. Similarly, for predicted program costs to outweigh benefits under both the group-only delivery and mixed-delivery models (implying a BCR less than 1), our early intervention impacts on cognition would need to nearly completely fade out, from 0.52 SD to 0.03 SD in the long term. Finally, scaling up the group-only version of Msingi Bora would have a negative ROI for the government if the average tax rate was 6.4%, which is less than half our calculated average tax rate.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. One-way sensitivity analysis by intervention delivery model.

https://doi.org/10.1371/journal.pmed.1003746.t006

In Table B in S2 Text, we simulate new estimates for ICERs, BCRs, and ROIs under the alternative “best case” cost scenario that assumes a more efficient scaled version of the program’s implementation, as well as that allows for socioemotional skills to further influence lifetime wages beyond the impacts of cognition. In the group-only arm, these adjustments increase our ICER to a 0.55 SD improvement in cognition per US$100 investment, with an associated BCR of 22.4 and an ROI of 227%.

Discussion

We conducted an economic evaluation of 2 potentially scalable and effective group-based ECD parenting interventions targeted to mothers with young children in rural Kenya. We found that the total costs per child were slightly smaller for the group-only model (US$140) than for the mixed-delivery model (US$145) that substituted periodic home visits for group sessions. Coupled with the fact that effect sizes in child outcomes were consistently higher for the group-only delivery model (although not statistically significantly so at traditional levels), these cost estimates imply that the group-only model was also the most cost-effective, with effect sizes per US$100 additional investment of 0.37 SD in cognition, 0.30 SD in receptive language, and 0.16 SD in socioemotional outcomes.

We also found that the long-term economic returns of scaling Msingi Bora could be extremely high if our predicted long-term program impacts hold. Focusing on cognitive outcomes, we find a BCR of 15.5 for the group-only delivery model and 10.6 for the mixed-delivery model, while the average ROIs for a government scaling the program are 127% and 55%, respectively. We would also require very extreme values of our input parameters to have negative ROIs and BCRs less than 1. For example, program costs would outweigh benefits under the group-only model with a discount rate of 17% or if our early impacts in cognition almost entirely fade out in the longer term from 0.52 SD to 0.03 SD.

The small difference in costs across delivery models of US$5 per child is perhaps surprising. However, it is worth remembering that in 12 of 16 sessions (75%), the delivery model was identical across intervention arms, so the only difference in costs comes from the 4 review sessions. In addition, even though CHVs in the mixed-delivery arm spent an average of 27 additional hours traveling to and delivering each review session as home visits, when priced at current local wage rates, the additional hours spent did not translate into large differences in monetary costs.

Our group-only delivery model remains highly cost-effective when we compare our ICERs based on a health provider’s perspective to those we calculate for 2 recent effectiveness studies from Pakistan and India reporting unit costs for similar parenting interventions targeted to children aged 0 to 3 from LMICs and focused on similar outcomes of children’s cognitive and language development [13,14]. We consider only provider costs simply because no earlier studies accounted for societal costs in their analyses, and we adjust all costs to 2020 US$ in purchasing power parity (PPP) terms to facilitate comparisons. Under this criteria, for Msingi Bora, we estimate an ICER of 0.176 SD improvement in cognition per PPP$100 investment. For the Lady Health Worker (LHW) Program from Pakistan, we estimate a lower ICER for cognitive outcomes of 0.148 SD improvement in cognition in similar 2020 US$PPP terms, based on a reported monthly cost per child of US$4.15 (in 2012 US$) in the responsive stimulation arm [13], which was the study arm with the largest impacts on cognition [37], and assuming that each child was treated for 24 months. However, for language outcomes, their program had a relatively larger ICER of 0.17 SD improvement per PPP$100 compared to our study’s estimated ICER of 0.14 SD improvement in language per PPP$100. It is worth noting that their program costs are at least partially lowered by the fact that their interventions were integrated within the existing infrastructure of government health services with already highly trained personnel, costs that were excluded from their analysis under a provider costing perspective. We imagine that our training costs, which are roughly one-third of our total provider costs, may be more realistic in similar settings where existing infrastructure and personnel are not on par with Pakistan’s LHW program [4]. Furthermore, their study does not account for private opportunity costs to participants or delivery agents during their 33-month program as we do.

Another recent effectiveness study conducted in Odisha, India compared group meetings versus individual home visits over 2 years and similarly found group visits to be much more cost-effective for a relatively similar sample to ours [14]. This study also did not account for private opportunity costs of participants or delivery agents. The total provider costs for group meetings were US$76 per child for the 2-year intervention, which translates to US$265 in 2020 PPP terms, implying a cost-effectiveness ratio of 0.106 SD improvement in cognition per PPP$100 investment and 0.114 SD in language. However, these reported costs per child are based on the assumed number of children that a scaled version of this intervention would be able to cover if facilitators worked full time, which is roughly 9 times higher than the actual number of children invited to the intervention. If we similarly assume that each CHV can work full time to conduct 8 group sessions per week, a scaled version of Msingi Bora using only group visits would cost PPP$89 per child using a societal perspective, where we assume a full-time CHV wage again based on local wage rates. At these assumed costs, we would estimate an ICER of 0.58 SD improvement in cognition per PPP$100 investment and 0.47 SD in language, which again makes our program much more cost-effective despite the inclusion of societal costs, which in a scaled version of Msingi Bora we estimate would account for 61% of the total cost per child. Besides having higher impacts on cognitive and language outcomes, other reasons our program is more cost-effective than the one conducted in India include that our group meetings were nearly twice as large as theirs (they invited 7 to 8 children while our median attendance was 13 children at group meetings), and while 23% of their total costs were allocated to purchase toys and learning materials, our program only used play materials freely available in the home.

Msingi Bora’s estimated BCRs are also comparatively high with the very limited number of related interventions reporting similar estimates (Table 7) including 3 home visitation programs in Latin America [32], an intervention delivered through health center meetings and home visits in Jamaica and other Caribbean countries [38], and a combined day care and home visitation program in Nicaragua [39], all of which used a lower discount rate (3%) than we did in our study (5%). Our BCRs are also high when compared to high-profile home visitation programs evaluated in the United States such as the Nurse-Family Partnership Program, for which recent evaluations report BCRs in the range 1.5 to 5 [40,41]. Furthermore, our BCRs and ROIs only consider the increase in discounted lifetime earnings as benefits of program participation and ignore any potential long-term impacts on other health and social outcomes, and, as such, can be interpreted as likely lower bounds on the true potential societal returns (under the assumption that the program impacts do not fade out in the long term).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Comparison of BCRs of Msingi Bora with similar ECD programs in LMICs.

https://doi.org/10.1371/journal.pmed.1003746.t007

Our study has numerous strengths related to methods and the economic evaluation itself. To the best of our knowledge, our study is the first economic evaluation of an ECD parenting intervention targeted to children under 3 years of age in SSA. Moreover, our study is unique in comparing the cost-effectiveness of different models of delivery contributing to the debate of potentially scalable models for ECD parenting interventions in low-resource settings. Strengths of our study include adoption of a societal perspective for our CEA that accounts for private opportunity costs to participants and delivery agents, the combination of cost-effectiveness and BCAs to assess outcomes, and the inclusion of sensitivity analyses showing the values at which the program generates negative returns. All of these complement each other to provide policymakers a more comprehensive portrait of private and social costs as well as expected economic returns of our program. Limitations of our analysis include that our BCA necessarily relies on different assumptions of how program impacts translate into long-term benefits and costs using external data sources and that we assume that our early impacts can be sustained in the long term. Future work could address these limitations, but obviously necessitate the passage of time. We also do not account for varying levels of schooling quality across public or private schools that may present different economic opportunities for children enrolled in Msingi Bora, and we do not account for patterns of formal and informal employment or internal migration that may influence the profiles of wages by age we used to estimate benefits. We also examine the economic benefits of enhanced human capital only in terms of increased wages, but we do not include other outcomes that could potentially improve such as labor market participation, share of workers in the informal sector, health, life expectancy, or prosocial behaviors. We may similarly be missing some societal costs in addition to those included. As discussed above, we believe the BCR outcomes can be interpreted as the lower bound of the true societal returns as long as our early impacts are sustained, but if there are large costs incurred to society as a result of intervention-related improvements, leaving these out may modify our BCR and ROI estimates.

Our findings highlight the need for careful cost analyses to accompany evaluations of ECD parenting programs in LMICs to help inform governments and policymakers on the potential returns to society and to the government of investing in such programs. The growing recognition worldwide about the importance of a child’s early years and the potential promise of responsive parenting interventions to improve children’s outcomes are not likely to affect policy decision-making without accompanying information about the cost-effectiveness of such investments. We also believe that cost-effectiveness analyses of ECD parenting programs should account for the private opportunity costs to participants and delivery agents for both ethical and practical reasons: Excluding these costs could create perverse incentives for governments to offload program costs onto users and delivery agents, who are often vulnerable and poor citizens whose burden of time should not be assumed to be free. We hope our results can motivate similar analyses from other ECD parenting programs from LMIC settings.

Our results show that an integrated child development intervention featuring exclusively group sessions delivered by paraprofessional community health workers can be very cost-effective. We find that a mixed-delivery model that substitutes periodic home visits for group sessions comes at a slightly higher cost, albeit with a lower benefit, and thus is less cost-effective for our setting. Our results represent a strong case for exploring pathways to scale similar interventions in low-resource rural settings where many children and families could potentially benefit from ECD parenting programming, and investments could pay significant dividends over time.

Supporting information

Acknowledgments

We thank Alie Eleveld, Alex Mwaki, Aloyce Odhiambo, Edith Alu, Ronald Ochieng, Nicholas Ogendo, the CHVs, and the entire team at the Safe Water and AIDS Project (SWAP) in Kenya for invaluable support in carrying out this study. We thank Michael Walker, Michelle Layvant, and the team at UC Berkeley for help generating the estimates for impacts on wages and schooling using KLPS data. We extend our gratitude to Frances Aboud and Rebecca Kilburn for providing feedback on earlier versions of this draft.