Digital technologies in enhancing hierarchical vision health management for the next 1000 days of children’s life: multi-component economic evaluation

Summary

Background

Visual impairment in the next 1000 days of children’s life presents heavy medical and economic burdens. It is urgent to bridge the gap between early intensive healthcare engagement and school-based services to ensure timely detection during the pre-disease period. Digital technologies have shown promise but lack for economic evidence interventions targeting early childhood. Thus, we proposed and evaluated the economic value of a full-process digital-empowered hierarchical vision management strategy (FDH strategy) for preschool-age children in China.

Methods

A decision-analytic Markov model was constructed to identify the cost-utility, cost-effectiveness and economic burdens among no intervention, traditional screening strategy, tele-screening strategy and FDH strategy for a hypothetical cohort which consists of 100,000 preschool children (3 years old). The parameters were obtained from published sources. Primary outcomes include incremental cost-utility ratios (ICURs), incremental cost-effectiveness ratios (ICERs), and net monetary benefit (NMB). ICURs were calculated using quality-adjusted life-years (QALYs), and ICERs using disability-adjusted life-years (DALYs). These outcomes were analyzed separately for rural and urban settings. Broad sensitivity analyses were performed to test the robustness of base-case analysis.

Findings

Among all strategies evaluated, the FDH strategy was the most economically attractive option. It produced an ICUR of $780 (95% CI: 130–912) and an ICER of $5753 (95% CI: 1615–6894) in rural settings; it also exhibited had the lowest cost per QALY gained and per DALY averted in urban settings. Similarly, the FDH strategy had the highest NMB, with values increasing from $75 to $10,207 (rural) and $146 to $11,267 (urban) over the 15-year period. In rural areas, the FDH strategy reduced total myopia prevalence by 26.51%, high myopia by 15.35%, and combined strabismus, amblyopia and hyperopia by 12.6%. In urban areas, the corresponding reductions were 19.94%, 15.03%, and 5.97%. By further comparison, without LLM-based health promotion, costs were higher for achieving health benefits, which resulted in a cost increase of $146 per QALY gained and $2214 per DALY averted in rural settings; in urban settings, the increase was $273 per QALY gained and $1321 per DALY averted. All the above results were robust according to the sensitivity analysis.

Interpretation

Multi-component health economics evaluations confirm the FDH strategy is highly cost-effective at a national level. This offers a valuable insight and economic rationale for promoting equitable, accessible, and sustainable eye health care during early childhood for countries with limited ophthalmic resources.

Funding

This study was supported by grants from National Natural Science Foundation of China (82171051), National Natural Science Foundation of China (82371053), Excellent Young Scientists Fund of the National Natural Science Foundation of China (82422018), Development and Reform Pilot Project 2025: Research on Molecular Age Prediction Model of Eye and Whole Body Organs Based on Multi-omics and Imaging, Beijing Municipal Public Welfare Development and Reform Pilot Project for Medical Research Institutes (PWD&RPP-MRI, JYY2023-6), and Sanming Project of Medicine in Shenzhen (SZSM202411009), Key Project of Beijing Natural Science Foundation for Young Scientists (JQ25021).

Research in context.

Evidence before this study

We conducted a literature search in PubMed, Science Direct, MEDLINE, and CNKI using the terms: “myopia,” “strabismus,” “amblyopia,” “esotropia,” “exotropia,” “eye diseases,” “children,” “students,” “ophthalmology,” “screening,” “treatment,” “health care,” “machine learning,” “artificial intelligence,” “deep learning,” “economic evaluation,” “cost-effectiveness analysis,” “cost-utility analysis,” “cost-benefit analysis,” “cost-minimization analysis,” “cost of illness analysis,” “disease burden,” “China,” and “Chinese” (alone or in combination), with publications restricted to those published before November 1, 2024. We also reviewed the references of identified studies. Nevertheless, current research on preschool vision health focuses mainly on traditional screening/treatment, with narrow scope: single-center, isolated interventions, or individual eye diseases. Early childhood visual impairment especially during the Next 1000 days brings heavy medical and economic burdens. Traditional approaches fail to meet demands for comprehensive, timely, and intensive vision care. Digital technologies are increasingly integrated into ophthalmology, alleviating the issue of limited ophthalmic resources to enable community/home-based scenarios, integrate multi vision care process and supports real-time assessment—showing promise for meeting the demands of preschool vision care. However, existing economic studies related to digital interventions are primarily targeted at age-related eye diseases, with limited attention to preschool children. Thus, assessing the economic value of nationwide digital vision management for preschoolers—especially in China’s ophthalmic resource-constrained settings—is essential, as such interventions could most reduce long-term visual impairment and societal costs.

Added value of this study

To our knowledge, this is the first national, multi-component economic evaluation of a full-process, digital hierarchical vision management strategy for children in the next 1000 days. By integrating multi-component economic evaluations—cost-of-illness, cost-effectiveness, and cost-utility analyses, we demonstrate that the FDH strategy is the most economically attractive to significantly improve the disease outcome, services coverage and cost savings, especially for rural areas. And it has the potential to significantly alleviate the economic burden from productivity loss in adulthood. Moreover, the integration of ophthalmic Large Language Models (LLMs) is crucial for reducing the medical cost and enhancing accessibility and adherence among children.

Implications of all the available evidence

The results of our study prove that the FDH strategy emerges as a cost-effective and scalable approach to bridge the eye care services cracks during the next 1000 days in real-world clinical environments. The findings of our study also offer new insights and robust evidence to promote equitable, accessible, and sustainable eye health for children, particularly for countries seeking affordable pathways to unleash children’s potential and foster sustainable socioeconomic development. Successful implementation of the FDH strategy hinges on seamless communication and collaboration among families, educators, and healthcare providers. Moreover, enhancing digital literacy among parents and teachers is crucial for the effective deployment of the FDH strategy.

Introduction

Globally, over 1 million children and adolescents suffered vision impairment (VI),¹ leading to an annual productivity loss of US$410.7 billion.² Worryingly, the economic burden of near vision loss doubled from 5.3 million DALYs in 1990 to 9.8 million in 2017 worldwide.³ In low- and middle-income countries (LMICs), VI accounts for half of the burden of childhood disability.⁴ The majority of eye diseases have not been identified during the first 1000 days of life from conception to two years of age. The leading common VI, including uncorrected refractive error (URE), amblyopia, and strabismus, are best managed via prevention and rehabilitation in the next 1000 days—a critical period of vision development in early childhood.^5,6 However, during this pre-disease period, many children fall through the cracks between early intensive healthcare engagement and formal school-based health services. This situation is exacerbated for children in remote areas due to limited ophthalmic resources.

Bridging the eye health crack in early childhood is crucial, as untreated early ophthalmic diseases can lead to irreversible VI or blindness. Childhood-onset VI not only affects vision but also motor, cognitive, and psychomotor development, impacting academic performance, careers, and life trajectory.^7,8 Childhood-onset VI poses a significant burden due to the economic and opportunity costs associated with years of vision loss, care expenses, and potential productivity losses. Thus, prioritizing vision health care during early childhood is essential for bridging cracks, particularly for primary eye care, which is vital for managing children’s visual health. However, the vast majority of eye care services are provided in secondary or tertiary hospitals, which are principally located in urban areas. The maturation of low-cost, scalable digital ophthalmic technologies now provides opportunity to enhance primary eye care services, while the economic value of these technologies in eye diseases has been well-documented, especially for the age-related eye diseases,9, 10, 11, 12 their potential to transform eye care for children during the early childhood remains unaddressed. Unlike adults, children’s vision health management is characterized by a long period, intensive human resource requirements, and wide societal impact. The potential of digital technologies to address eye health care during early childhood remains largely unexplored, creating significant evidence gaps for policymakers. Therefore, it is imperative to evaluate the economic value of eye care services empowered by digital technologies for children in early childhood at the national level before large scale implementation.

To address this critical evidence gap, we developed a nationwide Markov model integrating three complementary economic methodologies: cost-of-illness (COI), cost-effectiveness analysis (CEA), and cost-utility analysis (CUA) to evaluate the full-process digital-empowered hierarchical (FDH) vision management strategy for children during early childhood, which could offer a detailed road-map to formulate eye health policies to bridge the cracks for countries with limited ophthalmic resources.

Methods

Overview of model design

A Markov model was developed using TreeAge Pro (Williamstown, MA) to evaluate incremental cost-utility ratios (ICURs), cost-effectiveness ratios (ICERs), net monetary benefit (NMB), and COI of three vision health management strategies, as illustrated in Fig. 1: the traditional strategy (screening by ophthalmologists with ophthalmic examination equipment), the tele-screening strategy (screening by kindergarten health providers), the FDH strategy (AI screening by non-medical screeners, supported by a multi-digital technology platform that integrates deep learning, Internet of Things (IoT), Large Language Models (LLMs), big data, cloud storage, and 5G communication technology, with functions including screening, automatic referral and triage, electronic medical record generation and early warning, rehabilitation and self-monitoring, and health promotion).

Fig. 1.

The workflow of the traditional strategy (a),Tele-screeningstrategy (b) and FDH strategy (c).

The flowchart of vision management strategies evaluated in this study: (a) The traditional strategy uses ophthalmologist-led screening, referring suspected cases to secondary/tertiary eye care for diagnosis, treatment, and rehabilitation; non-referable children receive regular screening. (b) The tele-screening strategy involves kindergarten health providers screening with tele-ophthalmology support; suspected cases are referred to secondary/tertiary care under GP supervision, while non-referable cases remain in regular screening. (c) The FDH strategy expands screeners to non-medical screeners (parents, teachers and volunteers) supported by AI screening technology, across settings (children’s health services, kindergartens, communities, families, private optometry), and integrates multiple functional modules (screening, automatic referral and triage, electronic medical record generation and early warning, rehabilitation, self-monitoring, and health promotion) supported by multiple digital technologies (deep learning, Internet of Things, LLMs, big data, cloud storage, and 5G communication technology). Suspected cases, having been automatically referred and triaged by GPs with AI assistance, are sent to primary eye care for simple conditions, secondary care for complex treatment or rehabilitation, and tertiary care for surgery and training programs. Child-centered active eye health, including electronic medical records, self-eye health monitoring, health promotion, and refractive error services, interacts with the eye care system, which provides supervision and guidance.

To simulate real-world vision screening management—where children undergo screenings twice a year—the Markov model is structured with two cycles per year. It covers the period from age 3–18 years (a 15-year span), with a total of 30 cycles. Based on the leading VIs in this age group, children in the model were categorized as normal, myopia, abnormal hyperopia (SE ≥ 6.00 D), strabismus (esotropia/exotropia), and amblyopia (Supplementary Fig. S2). Myopia was categorized into low-to-moderate myopia (−6.00 D < SE ≤ −0.50 D) and high myopia (SE ≤ −6.00 D). Each stage has a defined probability of transitioning to the next (Supplementary Figs. S1 and S2). No ethics approval was required for this study.

To ensure the transparency and reproducibility of economic evaluation details of the FDH strategy—which relies on multi-digital technologies—the subsequent analyses were conducted in adherence to the Consolidated Health Economic Evaluation Reporting Standards for Artificial Intelligence-based Interventions (CHEERS-AI) checklist.^13,14

Data input

Primary parameters

Primary parameters include prevalence, transition probability, and compliance, were derived from real-world studies, meta-analyses, pilot projects and official data (National Health Commission of the People’s Republic of China, National Expert Advisory Committee on Vision Health Management for Children and Adolescents, National Committee for the Prevention of Blindness, and National Institute of Health Data Science at Peking University) and reasonable assumptions (Supplementary Tables S1–S4). The transition probability (r) between vision states over a half-year cycle was calculated using the formula: r = −[ln (1 − p)]/t, where p indicates the cumulative incidence over time interval t.¹⁵ The transition probability was age-specific, with distinct values calculated for each annual age interval.

Costs

According to the process of vision management, the expenses includes the screening, hospital diagnostic examination, treatment, rehabilitation, and health education. From a societal perspective, the semi-annual costs of vision management consisted of direct costs (medical and non-medical) and indirect costs.

Direct costs result from the expense of therapy and rehabilitation incurred inside and outside the hospital, as well as costs related to transportation, food, and accommodation associated with visits; indirect costs included family members’ working time loss and accompanying costs due to the visit to hospital. Both capital and recurrent costs were included in our study. All costs were collected in Chinese yuan and converted into US dollars at the average exchange rate in 2024 (CNY1 = US$ 0.1404). The composition of costs is shown in Supplementary Tables S5–S14.

Main outcomes

Main outcomes include ICURs, ICERs, NMB and COI. All the economic assessments were discounted at an annual rate of 3.5%.

ICURs, ICERs and NMB were calculated according to the following formulas:

ICURs = (incremental cost)/(quality-adjusted life-years gained).

ICERs = (incremental cost)/(disability adjusted life years averted).

NMB = (quality adjusted life years gained) × WTP − incremental cost.

In terms of the cost-effectiveness threshold, according to WHO,¹⁶ an intervention is considered not cost-effective if its cost-effectiveness exceeds 3 times the per capita GDP, and highly cost-effective if it is lower than 1 time the per capita GDP. The per-capita GDP for rural and urban areas in China was $12,013 and $13,851, respectively, inferred from the overall national per-capita GDP ($13,491), the urban–rural ratio (2.34) of per-capita disposable income, and the urbanization rate (0.67) in 2024. The WTP threshold in this study was (3 times the per-capita GDP) $36,039 in rural areas and $41,553 in urban, per quality-adjusted life-year (QALY) gained or per disability-adjusted life-year (DALY) gained. NMB quantifies the net economic value of the evaluated strategy, compared with no screening, indicating whether its monetized benefits outweigh the incremental costs incurred. A positive annual NMB signifies that the strategy’s monetized benefits offset its costs (Supplementary Tables S15 and S16).

The COI, referring to disease burden in this study, was calculated according to the following formulas:

COI = Medical costs + Productivity losses.

Productivity losses were stratified by VI severity:

For low-to-moderate VI (low-to-moderate myopia, hyperopia, or strabismus):

Productivity losses = Number of patients × Labor participation rate × Per capita annual income loss.¹⁷

For severe VI (high myopia, severe strabismus, or amblyopia):

Productivity losses = Number of patients × Labor withdrawal rate × Per capita annual income.¹⁷

To estimate the lifelong economic burden derived from childhood VI under the evaluated strategies, a population-level extrapolation was conducted. The Markov model simulated vision progression exclusively for ages 3–18, as the evaluated strategies are restricted to the included VI conditions, whose prevalence rates stabilize in adulthood with minimal risk of significant progression. The prevalence data is shown in Supplementary Fig. S3 and Table S17. For ages 3–18, only medical costs were included, since this group makes minimal contribution to social productivity. For ages 18–60, the burden encompassed both medical maintenance costs and VI-attributable productivity losses. Individuals over 60 were excluded, as they are beyond peak productivity and formal workforce participation. The details of calculation is presented in Supplementary Tables S18 and S19.

Sensitivity analyses

To assess the uncertainty and robustness of base-case scenarios, broad sensitivity analyses were conducted. For one-way sensitivity analysis, a floating range of 10% of basic values was set as upper and lower bounds for prevalence, utility, sensitivity, specificity, compliance, and transition rate. In terms of costs, a floating range of 50% of basic values was set as upper and lower bounds. In probabilistic sensitivity analysis (PSA), a beta distribution was applied to prevalence, compliance, utilities, and transition rate, and a gamma distribution was applied to cost parameters for 10,000 Monte Carlo simulations.^18,19 The results of the sensitivity analysis are shown in Supplementary Tables S1–S4, and S20. The 95% CIs for the ICURs and ICERs were calculated using the percentile-based non-parametric bootstrap method. Subgroup sensitivity analyses were conducted according to the integration of LLMs in health promotion.

Results adhering to CHEERS-AI are shown in Supplementary Table S21.

Role of the funding source

The funders of the study had no role in the study design, data collection, data analysis, data interpretation, drafting of the manuscript, or decision to publish.

Results

The base-case analysis is presented in Table 1. In both rural and urban settings, the ICURs and ICERs of the traditional, tele-screening, and FDH strategies indicated high cost-effectiveness compared with no screening. Of these, the FDH strategy was the most economically attractive: it produced an ICUR of $780 (95% CI: 130–912) and an ICER of $5753 (95% CI: 1615–6894) in rural settings. Similarly, in urban settings, it was a dominant strategy and exhibited the lowest cost per QALY gained and per DALY averted across all evaluated strategies.

Table 1.

Base-case results of cost-utility analysis.

	Per person			Per 100,000 people			ICUR (95% CI; $)	ICER (95% CI; $)	NMB
	Costs, $	QALYs	DALYs	Incremental costs, $	Incremental QALYs	Incremental DALYs averted
Rural setting
No screening	329	18.082	0.088	–	–	–	–	–	–
Traditional	828	18.243	0.061	49,900,000	16,139	3105	3091 (355–4608)	16,064 (8652–16,265)	4863
Tele-screening	541	18.289	0.057	21,200,000	20,697	2717	1024 (163–1124)	7804 (1766–8917)	6664
FDH strategy	574	18.396	0.046	24,500,000	31,460	4258	780 (130–912)a	5753 (1615–6894)a	10,207
Urban setting
No screening	1534	18.085	0.103	–	–	–	–
Traditional	1419	18.282	0.073	−11,500,000	19,771	3317	Dominating	Dominating	8119
Tele-screening	1,393	18.279	0.070	−14,100,000	19,440	3044	Dominating	Dominating	8011
FDH strategy	1260	18.357	0.059	−27,400,000	27,152	4425	Dominatinga	Dominatinga	11,267

Note: QALY = quality-adjusted life-year. ICUR = incremental cost-utility ratio. DALY = disability-adjusted life-year. ICER = incremental cost-effectiveness ratio. All cost-effectiveness analyses in this table are compared against the “no screening” scenario as the reference baseline. Costs are given in US dollars. Both direct and indirect costs were taken into consideration. The cost-effectiveness thresholds were $36,039 per QALY gained or per DALY averted for rural settings and $41,553 per QALY gained or per DALY averted for urban settings. Costs, QAL-Ys, and DALYs are defined as lifetime values (from 3 to 18 years old) per person, whereas incremental costs, incremental QALYs, ICURs, incremental DALYs averted, and ICERs are defined as values per 100,000 people. Negative ICURs/ICERs are considered “dominating”, meaning the strategy has lower costs and more QALYs gained or DALYs averted vs. no screening.

^aFDH strategy has the lowest cost per QALY gained or per DALY averted among all strategies.

Over the 15-year period, the NMB of the FDH strategy was higher than that of the tele-screening and traditional strategies in both rural and urban settings. It increased from $75 to $10,207 in rural settings and from $146 to $11,267 in urban settings. The costs of the FDH strategy could be offset by its monetized benefits from the initial year.

As shown in Supplementary Fig. S3 and Table S17, the FDH strategy reduced the total myopia prevalence (including both low-to-moderate and high myopia) by 26.51% in rural settings and 19.94% in urban setting areas. High myopia prevalence decreased by 15.35% in rural settings and 15.03% in urban settings. For combined strabismus, amblyopia and hyperopia, the reduction was 12.6% in rural areas and 5.97% in urban areas. Economic burden analysis showed that, from an individual perspective (extrapolated to Chinese preschool-age children), the FDH strategy can save up to $331.5 billion in costs and effectively avoid $310.22 billion in productivity losses over the life-cycle of this population, aged 3 to 60, compared to no intervention (Supplementary Tables S18 and S19).

To assess the cost-effectiveness of LLM-based health promotion, we conducted a subgroup analysis. As shown in Table 2, without LLM-based health promotion, costs were higher for achieving health benefits, which resulted in a cost increase of $146 per QALY gained and $2214 per DALY averted in rural settings; in urban settings, the increase was $273 per QALY gained and $1321 per DALY averted.

Table 2.

The subgroup analyses FDH strategy with or without ophthalmic LLM Health Promotion Package.

	Per person				ICURs, $	Per person		ICERs, $
	Cost, $	Incremental costs, $	QALYs	Incremental QALYs		DALYs	Incremental DALYs averted
Rural setting
No intervention	329	–	18.082	–		0.088	–
With LLM–based health promotion	574	245	18.396	0.314	780	0.046	0.043	5753
Without LLM–based health promotion	568	239	18.340	0.258	926	0.058	0.030	7967
Urban setting
No intervention	1534		18.085			0.103
With LLM–based health promotion	1260	−274	18.357	0.272	Dominatinga	0.059	0.044	Dominatinga
Without LLM–based health promotion	1377	−157	18.299	0.214	Dominating	0.071	0.032	Dominating

Note: QALY = quality-adjusted life-year. ICUR = incremental cost-utility ratio. DALY = disability-adjusted life-year. ICER = incremental cost-effectiveness ratio. All cost-effectiveness analyses in this table are compared against the “no screening” scenario as the reference baseline. Costs are given in US dollars. Both direct and indirect costs were taken into consideration. The cost-effectiveness thresholds were $36,039 per DALY averted for rural settings and $41,553 per DALY averted for urban settings. Negative ICURs/ICERs are considered “dominating”, meaning the strategy has lower costs and more QALYs gained or DALYs averted vs. no screening.

^aWith LLM–based health promotion, the FDH strategy has the lowest cost per QALY gained or per DALY averted.

One-way deterministic sensitivity analyses indicated that the primary outcomes remained robust to a wide range of parameter fluctuations, and the ICURs for all strategies consistently fell within 1 time the per-capita GDP for both rural and urban settings. Supplementary Fig. S4 presents the top five parameters that had the greatest impact on ICURs for each strategy. PSA demonstrated that the FDH strategy had a 100% probability of being cost-effective in both rural and urban settings across 10,000 simulations. Similarly, in the rural setting, both the traditional strategy and the tele-screening strategy had a probability of 100% being cost-effective, while in the urban setting, their probabilities were respectively 99.96% and 99.99% (Supplementary Fig. S5).

In this study, the acceptability curves based on cost-effectiveness indicate that the FDH strategy is the dominant strategy across 99.82% and 98.58% of simulations in rural and urban areas, respectively, under the current willingness-to-pay (WTP) threshold (Supplementary Fig. S6). Moreover, the expected value from the perfect information analysis of the FDH strategy for rural settings ($0.89) and urban settings ($24.05) are both at a relatively low level which aligns with the findings from the PAS that the FDH strategy was the optimal strategy for the various settings.

Discussion

To our knowledge, this constitutes the first national, multicomponent economic evaluation of a comprehensive, full-process, multi-disease digital empowered hierarchical vision management strategy targeting children during the critical next 1000 days period. It assesses both individual and societal perspectives, and rural vs. urban economic conditions. Results suggest the FDH strategy is most economically attractive under the current WTP threshold.

VI poses an increasingly significant public health challenge, necessitating urgent prioritization of comprehensive vision health management strategies to alleviate its growing disease burden. The health economic value of vision interventions has been widely proven. Only a handful of studies have evaluated the economic value of traditional vision care during early childhood, such as kindergarten-based screenings or related treatments.20, 21, 22, 23 However, these studies often overlook the complex interplay of demographic characteristics and epidemiological features that influence the economic value of vision interventions due to their limited scope, which focuses on single-center implementations, single interventions, or single diseases. For instance, amblyopia screening among preschool children through parental and teacher-led initiatives in China was demonstrated to be cost-effective.²³ Conversely, a study in Canada reached the opposite conclusion due to the low prevalence of the amblyopia.²⁰ It has also been confirmed that implementing combined screening programs for multiple eye diseases can be highly cost-effective.⁹ Moreover, digital technologies that facilitate cross-sector collaboration among families, educators, and healthcare providers within hierarchical systems are highly effective in enhancing the efficiency, quality, and resource utilization of children’s visual health. The application value of digital technology for the visual health of preschool children has been demonstrated.^24,25 While the cost-effectiveness of digital technology in managing age-related eye diseases (such as cataract, diabetic retinopathy, and glaucoma),9, 10, 11, 12 myopia in school-age children,²⁶ and retinopathy of prematurity during the first 1000 days of life^27,28 has been established, the economic value of vision management for the next 1000 days of children’s lives—a critical developmental stage where early intervention can yield lifelong benefits—remains largely unaddressed. This highlights the urgent need for further research in this area. This study innovatively proposes the FDH strategy, which aims to address this gap by providing robust economic evidence and actionable insights specifically for countries facing limited ophthalmic resources like China.

Early childhood represents a vulnerable yet critical period for children’s development, with profound impacts on their lifelong health and well-being trajectories. Neglecting eye health services during this time incurs substantial long-term costs.⁶ Our predictive model indicates that, in the absence of vision management, over 80% of urban children are projected to develop myopia, aligning with previous research,²⁹ while 13% of rural children face a risk of permanent, irreversible VI due to missed opportunities for timely detection of amblyopia/strabismus and loss of optimal rehabilitation windows. Investing in eye health during early childhood can yield enormous lifelong benefits for children and generate significant economic returns that far exceed the initial costs. As our results demonstrate, the FDH strategy substantially reduces the incidence of irreversible amblyopia and strabismus in rural areas to 0.08%. The FDH strategy could save approximately $331.5 billion, with 94% of these savings coming from reduced productivity losses. While the strategy requires greater healthcare spending in rural areas, the productivity gains it generates are 207 times larger than the initial investment. Moreover, the FDH strategy is highly cost-effective and dominates as the most economically attractive option, with the lowest cost per QALY gained and per DALY averted in both rural and urban settings. This multi-component economic analysis indicates that FDH holds promise for bridging the cracks in eye care services during early childhood for countries seeking cost-effective means to unleash children’s potential and promote sustainable socioeconomic growth, especially in LMICs.

During early childhood, it is urgent to increase investment in eye care services for preschool-age children in resource-limited areas where quality, affordable eye care is scarce. Globally, significant cross-regional disparities in economic development and healthcare resources contribute to the “Matthew effect,” exacerbating the challenges faced by vulnerable children in accessing equitable eye care services. It has been reported that the proportion of children who have ever undergone visual acuity screening during early childhood is approximately 30% in both China³⁰ and India.³¹ This proportion drops to as low as 3% in remote areas of China.²¹ For children living in undeveloped areas, the prevalence of spectacle optical correction is approximately 20%,32, 33, 34 with the figure dropping as low as 0% in the backward regions of Africa.³⁴ In China, 41% of left-behind children lose the opportunity of school-based vision screening due to not been enrolled in preschool.³⁵ Compared with traditional approach, digital technology is a cost-saving approach to balancing ophthalmic resource distribution. The FDH strategy harnesses digital advancements to overcome traditional barriers, ensuring vision management—no longer limited to specialist ophthalmologists or fixed settings—offers accessible eye care services for children during early childhood. Previous evaluations from China, the United States, India, and Kenya have confirmed that training non-ophthalmic screeners—including nurses,³⁶ teachers,²¹ general practitioners (GPs),^37,38 and parents²³—can effectively administer vision screenings to enable more children to benefit. Our results show that the FDH strategy, implemented by families, educators, and healthcare providers, significantly boosts eye health service coverage. In rural areas, coverage increased by 3.54-fold compared to no intervention. Thus, the FDH strategy holds significant potential in ensuring equity, affordability, and accessibility of eye care services for children in early childhood in real-world clinical settings, particularly in primary care environments where digital interventions are likely to have a substantial impact.

Building upon the economic validation of the FDH strategy, we further quantify the economic value added by integrating Large Language Models (LLMs) into clinical practice. LLMs amplify the impact of the FDH strategy through two mechanisms. On the one hand, the persistent communication deficits in hierarchical healthcare systems limit children’s opportunities to access and adhere to eye care services. LLMs have diagnostic and triage capabilities that are on par with or surpass those of junior ophthalmologists,39, 40, 41, 42, 43, 44 thereby boosting the efficiency of internal workflows in hierarchical healthcare systems. On the other hand, using LLMs in vision care shows great promise for delivering personalized and precise services.^39,45, 46, 47, 48 LLMs leverage cutting-edge natural language processing and machine learning to deliver personalized health education, consultations, and reminders based on children’s vision and medical data.^43,47 This enhances parents’ eye health knowledge and generates more high-quality and empathetic responses, thereby alleviating parents’ anxiety. Studies have shown that providing long-term health information reminders to parents can significantly lower the incidence of myopia.^49,50 In China, over 755 healthcare institutions have deployed LLMs, with over 500 achieving on-premises deployment (OPD) within their own infrastructure. The medical scenarios where they can be applied include: clinical services (e.g., pre-clinical assistance, decision support, and clinical documentation generation), hospital operations (e.g., hospital management and data management), personal health management services (such as chronic disease monitoring and health report analysis), and scientific research and education.⁵¹ The LLM-based health promotion seamlessly navigates diverse clinical scenarios across hierarchical healthcare, reducing labor costs and minimizing travel expenses resulting from cross-level referrals, while enhancing the adherence and effectiveness of vision management. As our subgroup analysis demonstrates that the integration of LLMs into the FDH strategy results in savings of $146 ($2214) and $273 ($1321) per QALY gained (per DALY averted) for rural and urban areas respectively. This provides a compelling economic rationale for policymaking in practical and low-cost childhood vision management for countries like China.

Children’s eye health is a fundamental pillar of high-quality social development, and strengthening early vision health management is imperative for formulating sustainable and equitable eye health policies. This study estimates the health economic value of the FDH strategy and demonstrates its application potential for early childhood. However, the national-level implementation might pose challenges and issues. Firstly, eye healthcare for preschool children requires seamless communication and collaboration among families, educators, and healthcare providers to complete the FDH strategy. Therefore, it is crucial to establish a collaborative management mechanism at the governmental level as advocated by the Lancet’s Early Childhood Development Series,⁶ ensuring joint implementation and achieving synergistic effects. Secondly, utilizing digital ophthalmic technologies require a high level of digital literacy, including data comprehension and technology application proficiency. However, rural guardians with low digital literacy struggle to adhere to standardized procedures, resulting in poor-quality self-reported data that weakens screening reliability over the long term. Sustaining such standardized operations grows even more challenging. It is imperative to establish reciprocal assistance and support mechanisms, conduct activities pertaining to digital education and training, and develop user-friendly software that is uncomplicated yet effective to empower families’ involvement in vision management. Thirdly, implementing the FDH strategy presents technical hurdles in ensuring multi-module stability. For instance, ensuring the scientific validity of recommendations generated by LLMs necessitates continual validation against dynamically updated clinical guidelines; otherwise, these recommendations may seem plausible but could be inaccurate.

There were several limitations in our study. Firstly, we comprehensively evaluated the common leading eye diseases in childhood, but other congenital eye diseases were not incorporated into our model. Meanwhile, routine healthcare for children involves not only vision health but also interdisciplinary health management, including physical examinations, management of obesity, evaluation of psychological distress, and dental care. Our study may lead to underestimated cost-effectiveness of FDH strategy in real-world settings. Secondly, the current cost of digital ophthalmic technology has not been standardized yet, and its marginal costs and benefits are impacted by the implementation scale. Thirdly, some parameters were estimated by experts within a reasonable range. However, the model proved robust and insensitive to wide parameter fluctuations through extensive sensitivity analyses. Lastly, as the evaluated strategies target VI during early childhood, and we did not conduct Markov simulations for adulthood, there is a risk of overestimating or underestimating the actual situation.

The multi-component, national-level economic evaluations reveal that the FDH strategy holds significant potential to address eye care cracks in the next 1000 days of childhood and offers a novel, cost-effective paradigm for promoting equitable, accessible, and sustainable eye health for children in countries with limited ophthalmology resources.

Contributors

Conception and design: Hanruo Liu and Ningli Wang.

Acquisition or interpretation of data: Shanshan Jin, Weiling Bai, Mayinuer Yusufu, Ruyue Li, and Kaiwen Zhang.

Drafting of the manuscript: Shanshan Jin.

Administrative, technical support: Fengju Zhang, Li Li, and Haidong Zou.

Critical revision of the manuscript for important intellectual content: Hanruo Liu and Ningli Wang.

Statistical analysis: Shanshan Jin.

Obtained funding: Hanruo Liu and Ningli Wang.

Supervision: Hanruo Liu and Ningli Wang accessed and verified the data. Hanruo Liu and Ningli Wang were responsible for the decision to submit the manuscript.

Data sharing statement

The parameters that we used in our model (text, tables, figures, models, and appendices) are available on reasonable request from the corresponding author (Hanruo Liu; hanruo.liu@hotmail.co.uk) under certain conditions (with the consent of all participating centers and with a signed data access agreement).

Editor note

The translated abstract in Chinese was submitted by the authors and we reproduce it as supplied. It has not been peer reviewed. Our editorial processes have only been applied to the original abstract in English, which should serve as reference for this manuscript.

Declaration of interests

The authors declare no conflicts of interest.