Cost and budget impact analysis of a school-based vision screening programme in Cambodia and Ghana: Implications for policy and programme scale-up

Highlights • Scale up of the model school-based vision screening programmes is feasible in low-income settings.• Study provides an incremental cost estimate for the implementation of school-based vision.• School Health Integrated Programming project issued technical guidance for policy-makers and planners.• Study identifies main cost drivers of a school-based vision programme.


Introduction
Around 1 billion people live with preventable or unaddressed vision impairment globally, including 123.7 million with unaddressed refractive error (URE) [1]. The burden disproportionally affects people living in low-and-middle-income countries (LMICs), underserved populations, and in rural communities.
URE and unaddressed presbyopia are the leading cause of moderate and severe visual impairment [1]. Refractive errors occur when there is a mismatch between the shape and length of the eye results in light focusing either in front or behind the retina causing blurred and/or distorted vision. URE can be easily diagnosed and treated with optical devices such as spectacles, contact lenses or refractive surgery. The global economic losses due to URE were estimated to be US$244 billion a year [2,3]. The epidemiology of refractive error in children is complex as prevalence and types of refractive error vary within and between countries [1,4,5]. The World Health Organization (WHO) and the Lancet Global Health Commission on Global Eye Health (2021) estimate that around 19 to 22 million children live with visual impairment globally, 12 million are due to URE, with the highest prevalence in South Asia, Southeast Asia and Western Sub-Saharan Africa, potentially negatively affecting quality of life, education opportunities and livelihoods of millions of children [1,6,7]. Surveys conducted in a number of LMICs showed that URE in children is an important public health problem with the proportion of visual impairment in children caused by URE ranging from 55% in Chile to 93% in China [8].
Children with poor vision face multiple barriers in education, as it is commonly accepted that a significant part of learning during the first 12 years of life occurs through vision. Schoolchildren with URE are at a major disadvantage in a range of classroom activities [9], which affects their school attendance and academic performance [10][11][12][13].
School Health and Nutrition (SHN) programmes use schools as platforms to deliver safe, simple, and effective interventions essential for child development and growth. These programmes provide an opportunity to address health needs (e.g. deworming provided by teachers, hearing screening etc.) and support the education goals of school age children by getting them into schools, keeping them in education, and improving their learning whilst there [14]. The third edition of the Disease Control Priorities Project (DCP3) recommends an essential package of interventions for child and adolescent health [15], which includes vision screening and correction of refractive error alongside deworming, insecticide-treated net promotion, key immunisations, oral health promotion, and school feeding with micronutrient fortification. The cost of vision screening and provision of ready-made spectacles in the essential package for LMICs is estimated at US$0.60 per child per year [16]. Evidence from mathematical simulation models also shows that annual vision screening in schools is a very costeffective intervention. Although the cost-effectiveness ratios vary depending on the population size, URE prevalence and school enrolment rates, school-based vision programmes were shown to be costeffective in all WHO regions with the cost per Disability Adjusted Life Year (DALY) averted ranging from I$67-130 in South-East Asia to I$ 458-734 in Europe [17,18].
Whilst return on investment and cost-effectiveness are important criteria for policy-makers to decide on which intervention to implement, this information does not necessarily provide an assessment of the affordability, particularly in resource-poor settings. The scarcity of real-world and context specific data in LMICs often inhibits governments from making informed choices about public funding allocations or effective management of publicly funded services. The lack of primary data on the costs and affordability of school health and nutrition interventions is a barrier to mainstreaming these activities within education sector plans and inhibits efforts to improve the efficiency and sustainability of government-funded programmes.
The purpose of this cost and budget impact analysis (BIA) was to generate evidence for policymakers to assess the costs and affordability of including vision screening into SHN interventions and for the delivery of these programmes at scale. The study was undertaken as part of the School Health Integrated Programming (SHIP) project which was delivered in 2016 in four LMICs: Cambodia, Ethiopia, Ghana, and Senegal. The aim of the SHIP project was to test the feasibility of integrating vision screening and refractive error correction within SHN plans and to build countries' capacities to implement these plans at scale. More specifically, this study calculates the incremental costs of integrated vision screening and considers the budget implications for programme scale-up at national level in two of these countries, Cambodia, and Ghana.

Study design
The SHIP project was managed by two international nongovernmental organisations (NGOs), Sightsavers and Partnership for Child Development (PCD), with oversight from the World Bank and funding from the Global Partnership for Education. In Ghana, activities were implemented by Sightsavers and PCD, in collaboration with the Ministry of Health and the Ministry of Education. Whereas in Cambodia, the Ministry of Education and Youth managed the funds and directly implemented the activities with the support from The Fred Hollows Foundation, Sightsavers, and Krousar Thmey, a Cambodian organisation. Data from the SHIP projects in Cambodia and Ghana was collected retrospectively to measure the incremental cost of implementing school-based vision screening as part of an existing school health programme. A budget impact model was developed to estimate the budget required for scaling-up the programme nationally in all public primary and lower secondary schools in both countries. The budget impact analysis follows the guidelines from the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) [19]. Both the cost and budget impact analysis were carried out using excel.

Description of the intervention
The project activities took place in-country between January and December 2016. The SHIP approach advocates for the delivery of basic screening by trained teachers using simple inexpensive screening kits and children are referred to eye care professionals when failing the vision screening test, or in the presence of any abnormal clinical signs or symptoms [20]. Near vision acuity is not specifically measured. However, near vision issues are screened (including hyperopia and binocularity) by asking children about symptoms of headaches and eye pain (related to near vision issues), as well as assessing the appearance of the eyes. A child with strabismus, amblyopia or near vision symptoms would be referred. All referred students are assessed by a mobile team of optometrists, and children who required further treatment were referred to the nearest eye care unit or facility. The mobile team is comprised of optometrists, who visit schools soon after the initial teacher screening. Ready-made spectacles are provided immediately after the refraction for children who need spectacles and who meet specific prescribing guidelines; for children with more complex prescriptions, custom-made spectacles are procured and delivered to the schools one to two weeks later. School-based vision screening by teachers have shown acceptable results in other contexts [19,[21][22][23][24]. Other studies concluded that vision screening by teachers lack the required specificity and/or sensitivity, even if authors generally concede that better teacher training or strict screening process could improve these results [25][26][27].

Study settings
In Cambodia, SHIP was implemented in two adjacent districts, Banthey Srey and Angkor Thum, in Siem Reap Province. The districts have a joint population of 67,108 people [28]. A total of 126 teachers were trained and 12,440 primary schoolchildren were screened. Among them, 72 (0.6%) received spectacles and 22 (0.2%) were referred for other eye treatments. In Ghana, vision screening was implemented in Denkyembour district in the Eastern Region, which has a population of 78,841 people [29]. A total of 120 teachers were trained and 10,099 primary and junior high schoolchildren were screened. Of these, 78 (0.8%) received spectacles and 235 (2.3%) were referred for other treatments. Both countries followed a similar delivery model described in the SHIP technical guidelines that are publicly available [20]. The only key difference in terms of activity was that in Cambodia the approach was already supported by the Ministry of Education and Youth and there were already trainers of teachers available and who required no further training.

Data collection
Data was collected retrospectively between January and March 2018. Project output data were extracted from the SHIP project reports. Additional data such as population demographics and school statistics (enrolment rates, numbers of students, teachers and schools), were obtained from the national Education Management Information Systems (EMIS) and existing national surveys.
Expenditures were collected from project partner accounting systems and financial records. The study did not include costs incurred by project beneficiaries (i.e. children and their families), which, given the project settings, were minimal. The project's budget included transportation allowances and the spectacles that were provided free of charge. The analysis was incremental and did not account for the costs of the available medical and school infrastructure and resources. Only expenditures related to the implementation of the programme in each country were considered. Also, the costs of transport to eye clin-ics for children requiring other treatments were covered by the programme and included in the analysis, but the costs of treatment itself were not (due to the lack of robust data).

Cost analysis
The unit costs for standard activities as per guidelines adopted by the SHIP programme have been calculated (Fig. 1).
Although the activities in the two countries were closely aligned with the guidelines, there were some deviations reflected in the unit cost of each activity (see Table 1).
For the activities that were recommended in the guidelines but not implemented in a given country (for example training of trainers in Cambodia), the standard costs were estimated based on the expenditure in the other country, adjusted for local price level using Purchasing Power Parity (PPP) ratios. Ready-made spectacles used in the SHIP Cross-cutting activities supporting the implementation:

Training of trainers
• Coordination of activities • Capacity strengthening Cost of spectacles project were donated. However, the value of donated spectacles was estimated using manufacturer prices and included in the standard costs.
The expenditures for the following activities have been included in the analysis: capacity and partnership strengthening (workshops); community sensitisation; coordination; training; screening at schools; treatment by mobile teams; spectacles; and referral to eye clinics (see appendix Table A). As we only consider incremental costs, salaries of teachers or optometrists have not been included as they would incur in the absence of the programme. However, per diems, transportation, and every other financial expenditure directly attributable to the programme have been incorporated in the cost analysis.
We used the average monthly market exchange rates for converting currencies, unless stated otherwise. All costs were then converted to 2020 US$ using inflation rates [30].

Budget impact analysis
The perspective adopted for the budget impact analysis is from the budget holder, that is the perspective of the government / Ministry of Education who will be taking over the implementation and scaling up of school-based vision screening in each country (Ministry of Education, Youth and Sport in Cambodia and the Ministry of Education in Ghana). Only the activities included in the standard SHIP package were considered for the BIA ( Table 1). The costs of any international staff used in the SHIP project were adjusted using the costs of local personnel based on standard government per diems and salary grids.
The targeted population is all public primary school and lower secondary schools' students and teachers of every provinces of Cambodia and every district of Ghana (Appendix Table C). Representing 94% and 72% of total primary school enrolment, and 98% and 85% of total secondary school enrolment respectively [31,32]. The number of children enrolled were projected using a flow model and following the UNESCO / International Institute for Educational Planning (IIEP) technical guidelines [33]. Parameters such as promotion, repetition, and dropout rates were assumed to remain constant and equal to those observed in the base year (2016/17). Growth rates in new entrants were extrapolated from previous years (2011-2016) using both exponential growth rate and least square growth rate approaches [34]. A similar approach was used to estimate the number of new schools and teachers for each year.
The time horizon for the budget impact analysis is five years, which corresponds to the implementation of the Education Sector Strategy (2019-2023) in Cambodia and is aligned with the Education Strategic Plan in Ghana (2018-2030). A time horizon of one to five years is commonly used in BIA [19]. Prices were kept constant over the time horizon and no discounting was applied to future financial flows in the model.
All the key parameters and assumptions used in the budget impact model are outlined in the appendix (Table B). One-way and multi-way deterministic sensitivity analyses were performed to investigate uncertainties around certain parameters of the BIA model [35,36]. For oneway sensitivity analysis, the value of one parameter was increased and decreased by 25% while holding other parameters constant, as seen in literature [37]. For the multi-way sensitivity analysis, the value of several selected parameters was simultaneously changed by 25% using all possible combinations.

Ethical considerations
The economic analysis was based exclusively on the review of secondary data and no primary data collection was required. Administrative approvals were obtained from the relevant Ministries in both countries prior to the study. All data management procedures ensured anonymity and confidentiality. Power Parity (PPP) ratios. b Cost for sensitisation includes radio messages, posters, and other activities for reaching out of school children and building awareness amongst school and community on the importance of vision screening and correcting refractive errors. c One kit per school, screening kit for teachers contains a visual chart (6/9 optotype), a three meter rope and recording forms. d Based on project expenditure and number of days worked by mobile refraction teams. e Ready-to-clip spectacles used in SHIP were donated by Essilor; spectacles can be purchased for a unit price of 2 USD. f Personnel involved in eye health programme as per SHIP guidelines: one programme manager and one technical manager (at central level), and one coordinator per administrative unit (either province or region).

Table 2
Annual cost of national school-based vision screening programmes in Cambodia and Ghana, projection by activity for 2019-2023 (in USD 2020).  Table 1 presents unit costs for a standard package in each country. The activities where unit costs were similar in Cambodia and Ghana were training of teachers; community sensitisation; screening kits; ready-made spectacles; and transportation for referrals. The unit costs that differed substantially were capacity strengthening ($76 per participant/day in Cambodia vs $396 in Ghana); mobile refraction teams ($315 per day in Cambodia and $255 in Ghana); custom-made spectacles ($6.5 per pair in Cambodia and $25 in Ghana); and coordination at the central ($10,296 per year in Cambodia and $38,400 in Ghana) and regional levels ($13,310 per province per year in Cambodia and $28,738 in Ghana).

Budget impact analysis
If the vision screening programmes are scaled up nationally, the overall resources required over five academic years (2018/19-2022/23) are projected to be $11,375,046 in Cambodia and $26,755,707 in Ghana ( Table 2).
The first year is the most expensive for both countries due to programme start-up expenditures, capacity building and training activities, as well as the requirement of the SHIP guidelines to screen all children in primary and lower secondary schools in year one. In subsequent years, the guidelines recommend screening new entrants, children in secondary school (every other year) and re-examine children who were given spectacles previously. As a result, the programme start-up costs required for scale up were estimated at $1,758,525 in Cambodia and $4,255,400 in Ghana. Cost of screening and refractive error correction in year one was estimated at $1,392,694 in Cambodia and $3,660,599 in Ghana. With a training session every two years, recurrent expenditures in the subsequent year in Cambodia amounted to $1,165,805 in year two and $2,455,529 in Ghana. Table 3 indicates the cost and output breakdown per province/district for a five-year programme. It is anticipated that scaling up the school-based vision screening programme based on SHIP guidelines will lead to 5.5 million children being screened in Cambodia, with 31,703 receiving spectacles and another 9,687 being referred for further examination or treatment. In Ghana, over 9 million school aged children will be screened, among whom 70,270 will receive spectacles and 66,667 referred to an eye care facility for further examination or treatment.

Sensitivity analysis
Figs. 2 and 3 show results of the univariate sensitivity analysis. Six model parameters significantly impact the budget when their base case value is increased or decreased by 25%, or changing training frequency from annual, every two years or every five years. These parameters are the same for Cambodia and Ghana, although the order of importance of these parameters varies slightly.
In Cambodia, changes in the training frequency, the cost per teacher trained, the number of schools covered by the mobile team each day, and the cost of the mobile refraction team per day produce the Table 3 Cost and outputs of national school-based vision screening programmes in Ghana and Cambodia, projections by district/ province for 2019-2023 (in USD 2020). largest impacts on the five-year base case budget estimate ($7,866,869-$14,852,621). This is followed by the total number of schools to be covered and the cost of coordination (at regional level) which influence the total programme cost (Fig. 2). In Ghana, changes in the training frequency, the cost per teacher, the number of schools covered by the mobile team per day, and the cost of the mobile team per day have a large impact on the budget estimate ($18,137,534-$35,223,159). This is followed by the total number of schools covered, and the cost of coordination per region (Fig. 3).
Based on the multivariate sensitivity analysis, the projected total cost over five years ranges from $5,961,380 to $19,490,120 in Cambo-dia (base case value: $11,375,046), and from $13,399,511 to $46,705,375 in Ghana (base case $26,746,857) ( Table 4 and Table 5). This means that in both countries, the total cost of the five-year programme could be about 50% lower than the estimated base case value in the best-case scenario and 75% higher in the worst-case scenario.

Discussion
Prior economic analyses of school-based vision screening programmes in LMICs rely primarily on information from secondary  sources or WHO-CHOICE standardized unit costs. This is one of the first studies to estimate costs using real-world data from two countries, Cambodia and Ghana. This study also carried out a budget impact analysis to assess the affordability of delivering these interventions at scale in the studied settings. The share of resources required for the actual screening and refractive error correction were relatively modest, supporting the conclusions of earlier research that vision screening for schoolchildren by teachers is effective and less costly than alternative primary eye care models [38].
In this study, we did not have access to students' personal records on disability or special needs status. We are not aware of any reported cases where the standard screening procedure was not applicable to certain children. However, we acknowledge that it may be the case for children with some severe disabilities. We suggest that children unable to be tested using the screening protocol should be referred to optometrists who are better suited to examine complex cases, including children with special needs. Alternative ways of screening are also available and could eventually be added. For example Gogate et al., in India, used Kay pictures as an alternative to Snellen's tumbling E chart for young children and children with severe learning disabilities [39].
The standard unit cost based on the SHIP technical guidelines make cross-country estimates more comparable and highlight a number of context-specific differences which need to be taken into account when planning similar interventions in other settings. Except for the cost of the mobile team, the unit costs for all activities were higher in Ghana. These differences are likely to be due to different salary levels, per diem rates and accommodation, and rental costs. The results show the importance of using context-specific data when planning and budgeting interventions at scale.
It is difficult to place our estimates and projections in the context of other studies, as evidence on cost of school-based vision screening is scarce and the differences in methodological approaches of the few studies available limit opportunities for meaningful comparisons. In addition, this analysis did not aim to calculate the number of DALYs averted by correction of refractive error. Therefore, at this stage it is not possible to do a comparison with the WHO cost-effectiveness thresholds [40] or cost per DALY estimates presented in earlier studies [17,18]. The only estimates with which we can compare our results are from Baltussen, Naus & Limburg (2009), who reported the cost per child treated for refractive error. However, their estimates were made for children aged 11 + years and are available for the WHO Africa region only. They find that the cost per child treated for refractive error is I$204, when screening children aged 11-15 years and I $450, when screening 13-year-olds only [17]. When converting our results into the same currency (international dollars with the same base year), we obtain estimates of I$802 for Cambodia and I$726 for Ghana (dividing the projected number of children provided with spectacles by the total programme cost).
Our analysis also highlighted several programmatic aspects of vision screening which should be considered when designing and implementing similar programmes as they could potentially help reduce costs and enhance the benefits of school-based programmes. Firstly, most children with refractive error identified by the SHIP project were eligible for ready-made spectacles. Considering the price difference between ready-made and custom-made spectacles ($2 vs $6.5 in Cambodia and $2 vs $25 in Ghana), the use of ready-made spectacles provided on site should be prioritised, as it can generate substantial savings. Secondly, the use of teachers for conducting the initial vision screening is essential, as it significantly reduces the cost of the overall programme. Thirdly, the costs of activities and inputs shown  to be the key cost drivers, such as stakeholder mobilisation or teacher training, can be minimised if these activities are delivered as part of other planned programmes or training activities. Fourthly, as a large proportion of the costs required at the start of the vision screening programme are fixed, the delivery of the programme at scale would maximise the number of beneficiaries and reduce the unit cost per output achieved over the life of the programme. Finally, considering that teacher training and coordination constituted a significant part of the overall budget of the programmes at scale, we recommend further integration of vision screening with other school-based interventions to generate economies of scope. For example, training of teachers could cover other essential interventions, such as hearing screening, deworming, and health education [16].
One of the main purposes of conducting a budget impact analysis is to assess the affordability of a new intervention in a specific setting. The total budgets required to implement the national school-based vision screening programmes in Cambodia and Ghana were $11,375,046 and $26,755,707 over five years (2019-2023) respectively. It is therefore important to put the projected budget estimates in perspective with the education sector expenditure in each country.
In Cambodia, the national education budget in 2018 was equivalent to $848 million (Nary, 2017; Sokhean, 2017; Sophirom, 2017). In Ghana, it was $2.074 billion [41]. The delivery of the schoolbased vision screening programmes at scale is judged to be affordable in both countries, as the estimated five-year budget required for school-based vision screening would represent only 1.34%, 0.70% in the best-case scenario or 2.30% in the worst-case scenario, of the current annual education budget in Cambodia. In Ghana, it would represent 1.29%, 0.65% in the best-case scenario or 2.25% in the worst-case scenario of the annual budget. If we assume that the annual education budget remains constant during the 5-year period in both countries, the overall budget for the programme would represent only 0.27% of the five-year education budget in Cambodia and 0.26% in Ghana.
This study has several limitations which need to be considered when interpreting and using these findings. Firstly, the budget impact analysis is based on an important assumption that eye care services and skilled eye health personnel are available and sufficient for implementing the programme nationally. This assumption may not hold for all provinces or regions of the two countries. If the scale up of the vision screening programme requires investments in new infrastructure and human resources, the budget implications will be considerable. Secondly, the impact of COVID-19 has not been explored in our projections. The SHIP methodology is currently being adapted to embed the necessary social distancing measures, personnel protective equipment for optometrists, and adapted training sessions. This will have a financial impact on the implementation of school-based vision screening that has not been assessed yet. Moreover, national education budgets might be affected by the consequent economic recession in Cambodia and Ghana [42,43]. Thirdly, there are limitations regarding the availability of data used in the projection model including the prevalence of visual impairment and refractive error in school-age children. Education data was also unavailable for some administrative regions and the national averages were applied in the model. Moreover, data on sensitivity and specificity of teacher's screening ability, following the SHIP training, are still missing. The current literature show evidence supporting the approach of using teachers as first screener, but other examples show over-referrals, or even in some case under referrals, depending on the level of training of teachers. Fourthly, our analysis estimates the budget impact of integrating vision screening in government schools only and does not account for children enrolled in private schools or out-of-school. We did not have access to information on private schools in the countries. As private schools tend to cater for the needs of wealthier social groups, children from such families are more likely to access RE services through the private sector and it is unlikely that the governments in lowincome settings would be providing these services. However, the cal-culations for budget impact analysis suggested in this paper can also be used by the private sector or private insurance schemes, should such services be provided at scale by the private sector education providers. Finally, this study estimated the incremental cost of integrating vision screening into the existing SHN programmes and should not be interpreted as the true economic cost of the intervention.

Conclusion
This study suggests that the scale up of school-based vision screening programmes in resource limited settings, such as Cambodia and Ghana, is affordable for the current education budgets, providing there is sufficient in-country capacity to deliver such interventions at scale. The study highlights several policy and programme implications and provides suggestions for minimising costs and maximising efficiencies of vision screening in a school setting. Findings from this analysis can help education planners and international partners to improve their planning and budgeting processes for school-based interventions to improve health and learning outcomes for children in low-and middle-income countries.