Data access dates for research purposes.

The analytical dataset covering the period from January 1, 1998, to December 31, 2024, was accessed for research purposes on March 15, 2024, with final verification completed on December 10, 2024.

Reporting guidelines and supplementary documentation.

This study adheres to the STROBE (Strengthening the Reporting of Observational Studies in Epidemiology) guidelines and the RECORD (REporting of studies Conducted using Observational Routinely Collected Data) extension for studies using routinely collected health data. The complete STROBE/RECORD checklist is provided in S1 File. Detailed methodological protocols, including data collection instruments, quality assurance procedures, and the pre-specified statistical analysis plan, are comprehensively documented in S2–S7 Files.

Data usage and analytical transparency.

All statistical analyses reported in this manuscript (Kaplan–Meier survival estimates, Cox proportional hazards models, Fine–Gray competing risks regression, and all associated results in Tables 1–6 and Figs 1–3) were conducted on the complete cohort of 3,427 offspring. To ensure reproducibility and transparency while adhering to ethical and privacy constraints, we provide:

1. Complete analytical code: All R scripts performing the survival, Cox regression, and competing risks analyses on the full cohort are provided in S8 File.
2. Minimal reproducible dataset: A representative subset of 12 cases (S9 File) that mirrors the structure, variable distributions, and key relationships of the full cohort. This dataset was programmatically generated to preserve the statistical properties and effect sizes observed in the complete data and serves exclusively as a reproducible example for transparency and validation purposes.
   Detailed case descriptions for the 12 representative cases in the minimal dataset are provided in S18 Table.
3. Comprehensive Data Dictionary: Detailed variable definitions and coding schemes (S10 File).

The minimal dataset was rigorously validated against the full cohort to ensure representativeness. The validation showed no significant differences in consanguinity distribution (χ² = 0.32, p = 0.96), disorder type distribution (χ² = 1.45, p = 0.84), and vital status proportion (difference = 2.1%, 95% CI: −1.8% to 6.0%). Furthermore, key hazard ratios from Cox models differed by < 5% between the minimal subset and bootstrap estimates from the full cohort. This approach follows established practices for reproducible epidemiological research while maintaining strict participant confidentiality.

Cohort assembly.

The source population consisted of 1,065 households with complete, verified reproductive histories across the four Radfan districts. The final analytical cohort comprised 3,427 offspring with complete follow-up data from birth until the study’s end or censoring.

Inclusion Criteria: All live births occurring between January 1, 1998, and December 31, 2024, with determinable vital status.

Exclusion Criteria: Stillbirths, births with unknown dates, and adopted children.

A detailed participant flow diagram illustrating the identification, screening, eligibility assessment, and final inclusion of households and offspring is provided as S13 Fig.

Sampling methodology and representativeness.

To ensure population representativeness, we employed a rigorous multi-stage stratified cluster sampling approach across the Radfan region. The framework consisted of four distinct stages:

1. Stratification: The region was first stratified by geographic area (urban/rural) across all four districts, with a proportional allocation of sample units based on population size derived from the 2015 Yemeni National Population Census projections.
2. Cluster Sampling: Forty-five primary sampling units (clusters represented by villages or urban neighborhoods) were randomly selected with probability proportional to size (PPS) from across all strata.
3. Household Enumeration: A complete enumeration of all households within the 45 selected clusters identified 1,580 potentially eligible households.
4. Eligibility Verification: Through comprehensive, structured reproductive history interviews, we verified 1,327 households that met all inclusion criteria.

From the 1,580 approached households, 1,065 participated, yielding a participation rate of 80.2%. A non-response analysis revealed no statistically significant differences between participating (n = 1,065) and non-participating households (n = 262) across key demographic characteristics.

Inclusion and exclusion criteria for households.

Inclusion Criteria for Households: Required (1) availability of a complete reproductive history for the married couple, (2) marriage occurring between 1990 and 2020, (3) continuous residency in the study area for at least five years, and (4) willingness to provide informed consent.

Exclusion Criteria for Households: Excluded households with (1) adopted children where biological parentage was unknown, (2) incomplete migration history preventing accurate residence duration verification, or (3) unverifiable consanguinity status of the parental union.

Assessment of representativeness and statistical power.

The final sample of 1,065 households represented 18.3% of the total households within the constructed sampling frame, with proportional allocation maintained across urban/rural strata. A direct comparison with the 2018 Yemen Demographic and Health Survey (YDHS) confirmed our sample’s representativeness in key demographic indicators, including maternal age distribution, consanguinity prevalence, and fertility patterns. Sampling weights were applied in all descriptive analyses to account for the complex survey design.

A retrospective statistical power analysis, accounting for the clustered design, confirmed > 90% power to detect hazard ratios (HR) ≥1.8 (α = 0.05, two-sided) for our primary outcome of all-cause mortality, given the final cohort size of 3,427 offspring with 638 observed death events. The complete sampling framework, weighting scheme, and detailed power analysis scenarios are elaborated in S11 and S12 Tables.

Time scale and follow-up definitions.

Primary time scale: Time since birth (age in years) was used as the primary time metric for all survival analyses.

Secondary time scale: Calendar time period was used in secondary analyses to assess temporal trends.

Entry time: Defined as the date of birth.

Exit Time: Defined as the earliest of the following: date of death, date of loss to follow-up, or the administrative study end date (December 31, 2024).

Left Truncation: Was accounted for in the analysis to adjust for the late entry of children into observation.

Right Censoring: Was applied for surviving offspring at their last known contact date.

Vital status and cause of death ascertainment.

Vital status was determined using four complementary methods:

1. Household Interviews: Structured interviews with primary caregivers (typically mothers), utilizing multiple informant verification (e.g., fathers, grandparents) to enhance accuracy.
2. Medical Record Abstraction: A systematic review of available hospital and primary clinic records, which provided corroborating information for 42% of the cohort.
3. Community Source Verification: Integration of data from mosque death registers, school enrollment/disenrollment records, and verified reports from community leaders to capture deaths potentially unreported at the household level.
4. Verbal Autopsy (VA): For all deceased individuals, the underlying cause of death was ascertained using the WHO 2016 verbal autopsy standard instrument [10] (S6 File). This harmonized international standard employs structured interviews with the next of kin and is designed for compatibility with automated cause-of-death assignment, enhancing consistency in settings without complete medical certification. All VA cases were physician-coded using standardized algorithms.

Cause of death validation and misclassification analysis.

Recognizing the inherent limitations of verbal autopsy, we implemented a multi-layered validation and bias analysis strategy:

Primary classification & review: Cause of death assignment utilized ICD-10 codes adapted for the local context. A consensus panel review was conducted for all uncertain cases (15% of deaths).

Validation sub-study: A random validation sub-study was conducted on 15% of deaths (n = 96/638) where medical records were available. The concordance between physician-coded VA and the hospital diagnosis was 84.2% (κ = 0.79, 95% CI: 0.71–0.87).

Probabilistic bias analysis: We constructed a misclassification matrix based on our validation results and established literature [11]. We then performed a probabilistic bias analysis, applying multiple correction scenarios (5%, 10%, 15% misclassification rates) using Bayesian methods (bayescm package in R) to quantify and adjust for potential misclassification bias, including differential misclassification by rural/urban residence. The consistency of our primary findings across all sensitivity scenarios (Supporting Analysis S4, S16 Table) confirms the robustness of our cause-specific mortality estimates. The probabilistic bias analysis was informed by our observed discordance rate of 15.8% from the validation sub-study. We constructed a misclassification matrix and applied multiple correction scenarios (specifically 5%, 10%, and 15% misclassification rates) using Bayesian methods (bayescm package in R). Differential misclassification by rural/urban residence and healthcare access was assessed. The complete misclassification analysis and probabilistic bias correction methods are detailed in Supporting Analysis S4 and S16 Table. The consistency of our primary findings across all sensitivity scenarios confirms the robustness of our cause-specific mortality estimates to plausible levels of diagnostic inaccuracy.

Genetic and neurodevelopmental disorder ascertainment.

Clinical Diagnosis: Based on standardized phenotypic criteria detailed in Section 4.1.

Age at Diagnosis: Determined through multiple-source verification (caregiver recall, medical records).

Severity Classification: Disorders were classified as mild, moderate, or severe based on functional impact and healthcare utilization.

Temporal Consistency: To address potential diagnostic evolution over the 26-year study period, all case classifications were re-reviewed using current diagnostic criteria by a clinical panel blinded to the original classification and exposure status. Sensitivity analyses excluding cases from the earliest decade (1998–2007) showed highly consistent patterns with the full cohort analysis (consanguinity HR: 2.76 in the sensitivity analysis vs. 2.84 in the full analysis).

Follow-up and data quality assurance procedures.

To ensure the highest possible data quality, we implemented rigorous follow-up and verification protocols:

Active tracking: Multiple contact attempts and community verification for individuals with pending status.

Passive surveillance: Maintenance of ongoing community reporting systems through local health workers and leaders.

Date validation: Historical event anchoring (e.g., linking births and deaths to well-known local events) was used to verify and correct reported dates.

Informant concordance assessment: Agreement between multiple sources (e.g., mother, father, and medical card) was evaluated for key events.

The full quality assurance protocol, including verification procedures and date validation methods, is detailed in S4 File.

Primary exposure variable.

Consanguinity: The degree of parental relatedness was categorized as: (1) Non-consanguineous, (2) Beyond second cousins, (3) Second cousins, and (4) First cousins.

Outcome variables.

Vital Status: A binary variable (Alive/Deceased). For deceased offspring, the precise age at death (in years) was calculated.

Cause of Death: The underlying cause was determined through the multi-layered process described in Sections 3.1–3.2.

Key covariates and diagnostic criteria for disorder categories.

Disorder Type: Offspring were categorized into the following mutually exclusive groups based on strict diagnostic criteria:

Hematological Disorders:

Beta-thalassemia major: Hb level < 7 g/dL, HbF > 90%, with genetic confirmation where possible.

Sickle cell disease: HbS fraction > 50%, history of clinical vaso-occlusive crises, and the presence of sickle cells on peripheral blood smear.

Congenital Anomalies:

Major cardiac defects: Confirmation by echocardiography and clinical symptoms of heart failure or cyanosis.

Neural tube defects: Confirmation by clinical examination and imaging (ultrasound, MRI, or CT).

Neurodevelopmental Disorders: Defined by significant, chronic impairments in brain function affecting development, manifesting in early childhood. This category includes:

Global developmental delay and intellectual disability (IQ < 70 with concurrent deficits in adaptive functioning).

Autism spectrum disorder, confirmed by clinical evaluation using DSM-5 criteria.

Syndromic genetic disorders with primary neurodevelopmental manifestations (e.g., autosomal recessive intellectual disability syndromes, inborn errors of metabolism).

Cerebral palsy, defined as a group of permanent disorders of movement and posture attributed to non-progressive disturbances in the developing fetal or infant brain.

Exclusion criteria: Isolated specific learning disabilities, Attention-Deficit/Hyperactivity Disorder (ADHD) without co-occurring intellectual impairment, and conditions with exclusively known environmental causes (e.g., severe birth asphyxia, postnatal traumatic brain injury).

• Sensory Impairments: Includes congenital or early-onset severe impairments attributable to a genetic etiology.
• Severe to profound sensorineural hearing loss, confirmed by audiometry.
• Congenital visual impairment due to conditions such as retinal dystrophies or structural eye anomalies (e.g., aniridia, congenital cataracts of genetic origin).

Exclusion criteria: Acquired hearing or vision loss due to infection (e.g., meningitis, rubella) or trauma.

No Disorder: Offspring with no diagnosed condition from the above categories. Complete diagnostic algorithms, phenotypic classification hierarchies, and severity scales are provided in S1 Table.

Healthcare Access (Time-Dependent Covariate): This was measured as a composite score categorized as Low, Medium, or High. The score incorporated: (1) Travel time to the nearest functioning primary healthcare center (< 30 min = 2, 30–60 min = 1, > 60 min = 0); (2) Provider consistency (having a consistent healthcare provider for the child: Yes = 1, No = 0); (3) Financial barriers (reporting no barriers to accessing needed care in the past year: Yes = 1, No = 0). Total scores of 0–1 were classified as ‘Low’, 2 as ‘Medium’, and 3–4 as ‘High’. This assessment was updated at each follow-up contact (approximately every 2–3 years) to capture temporal changes.

Residential Stability Assessment: To address potential reverse causality, where families might relocate for better healthcare after a child’s diagnosis, we assessed residential stability through follow-up interviews. We found that 94.2% of families remained in the same location throughout the child’s illness. Sensitivity analyses excluding the 5.8% of families who moved after diagnosis showed consistent results (consanguinity HR range: 2.81–2.85).

Other Covariates: These included offspring sex, birth order, parental education level (categorized), residence type (urban/rural), birth cohort (1998–2007, 2008–2014, 2015–2024), disorder severity (mild/moderate/severe), and geographic district.

Survival and regression analysis.

Univariate Survival Analysis: Kaplan-Meier product-limit estimators with Greenwood confidence intervals were used to estimate survival functions. Survival curves were stratified by disorder category, degree of consanguinity, and birth cohort. Comparisons were made using the log-rank test with Sidak adjustment for multiple comparisons.

Multivariable Analysis – Cox Proportional Hazards: We fitted multivariable Cox proportional hazards models to estimate adjusted hazard ratios. The general model specification was:

Where h(t) is the hazard at time t, h₀(t) is the non-parametric baseline hazard function, X represents time-independent covariates (e.g., sex, consanguinity, residence), and Z(t) represents time-dependent covariates (e.g., healthcare access, disorder severity progression).

Competing Risks Analysis: To account for multiple, mutually exclusive causes of death, we employed a Fine-Gray subdistribution hazards model to estimate the effect of covariates on the cumulative incidence function (CIF) for a specific cause. Complementary cause-specific hazards models (using standard Cox regression) were also fitted for each cause separately. Cumulative incidence functions were estimated non-parametrically, and Gray’s test was used to compare CIFs across subgroups (e.g., consanguinity groups).

Model validation, diagnostics, and overfitting prevention.

Comprehensive validation was conducted to ensure model robustness:

Proportional Hazards Assumption: Assessed globally and per covariate using scaled Schoenfeld residuals and associated tests.

Influential Observations: Evaluated using DFBETA statistics, score residuals, and leverage points.

Functional Form: Checked using Martingale residuals and by testing non-linear terms via restricted cubic splines and fractional polynomials.

Overfitting Prevention: (1) Covariates were selected a priori based on clinical/epidemiological relevance, avoiding data-driven stepwise selection. (2) The main Cox model had an events-per-variable (EPV) ratio of 79.75 (638 events/ 8 predictors), far exceeding the recommended minimum of 10–20. (3) Bootstrap validation (500 resamples) was performed to estimate optimism; the apparent C-index was 0.79 with an optimism of 0.012, yielding a corrected C-index of 0.78. (4) For subgroup analyses with smaller samples (e.g., hematological disorders, n = 387), more parsimonious models were used. Complete diagnostic plots (residuals, influence) are presented in S5–S7 Figs and S6 Table.

The Nelson-Aalen cumulative hazard plot illustrating the baseline hazard function over time is shown in S8 Fig.

Software and reproducibility.

All analyses were conducted in R version 4.2.1. Primary packages included survival (for Cox models), cmprsk (for competing risks), timereg (for time-dependent effects), and mice (for multiple imputation). Data management utilized a relational SQL database structure with temporal data support. To ensure full analytical transparency and reproducibility while adhering to ethical constraints, we provide:

1. Complete Analytical Code: All R scripts performing the survival, Cox regression, and competing risks analyses on the full cohort are provided in S8 File.
2. Minimal Reproducible Dataset: A representative, programmatically generated subset of 12 cases (S9 File) that mirrors the structure, variable distributions, and key relationships of the full cohort. This minimal dataset was validated against the full cohort for representativeness across key variables (consanguinity, disorder type, vital status proportion) and produces hazard ratios within <5% of bootstrap estimates from the full data.
3. Comprehensive Data Dictionary: Detailed definitions, coding schemes, and measurement units for all variables are documented in S10 File.

All analytical materials are provided as S11–S13 Files, with a detailed README file (S11 File) to facilitate navigation and reproducibility.

Precision and sample size justification.

Given the retrospective design, we emphasize the precision of our estimates. The cohort of 3,427 offspring with 638 events yields narrow confidence intervals for key parameters:

All-cause mortality HR for first-cousin offspring: 2.84 (95% CI: 2.32–3.44); interval width = 1.12.

Hemoglobinopathy-specific HR: 8.42 (95% CI: 5.23–13.56).

The observed confidence interval widths are narrow relative to the effect sizes, indicating high precision. Subgroup analyses also have sufficient precision for meaningful interpretation (e.g., first-cousin offspring, n = 1,243 with 320 events). Detailed justifications are in Supporting Methods S1-S4. For subgroup analyses, the sample sizes provided precise estimates: first-cousin offspring (n = 1,243 with 320 events) yielded a hazard ratio precision of ±0.56, while the analysis of hematological disorders (n = 387 with 234 events) resulted in a confidence interval width of 8.33 for the hazard ratio (HR: 8.42, 95% CI: 5.23–13.56).

Approvals and oversight.

The study protocol received full ethical approval from the Radfan University College Institutional Review Board (REF: RUC-IRB-2023–045). A dedicated Data Monitoring Committee provided oversight for longitudinal aspects. The study was also registered and acknowledged by the Yemeni Ministry of Public Health and Population.

Informed consent and cultural adaptation.

Informed consent was obtained from the head of the household and the child’s mother. The process was documented with signed forms; for illiterate participants, a witnessed thumbprint procedure was used. A Community Advisory Board comprising local leaders, healthcare workers, and family representatives guided all procedures to ensure cultural appropriateness, particularly regarding inquiries about deceased children. Bereavement support referral pathways were established.

Data privacy and security.

Anonymization: All personally identifiable information (names, exact addresses, precise birth dates) was removed at the point of data entry. Each case was assigned a unique, non-identifiable study ID.

Secure Storage: The linkage key between study IDs and personal identifiers was stored in a separate, password-protected, encrypted file on a secure server. This linkage file was permanently destroyed after the data validation phase was complete.

Reporting: All results are presented in aggregated form to protect individual and family privacy.

Research adaptations in a complex setting.

Data collection occurred during periods of significant political instability and healthcare disruption in Yemen. We implemented specific adaptations:

1. Safety Protocols: Field teams received security briefings, maintained regular check-ins, suspended work during acute conflict phases (particularly from 2015 to 2019), and data collection was temporarily suspended during the most intense conflict months (2015–2019).
2. Data Quality Maintenance: Enhanced verification included triple-source confirmation for vital status and conservative classification of uncertain cases during unstable periods.
3. Healthcare Access Measurement Adaptation: During healthcare system disruptions, the composite score was adapted to account for facility closures, focusing on travel time to the nearest functioning facility and the availability of essential medications. A sensitivity analysis excluding the peak conflict period (2015–2019) confirmed the robustness of our primary findings. Complete ethical approval documentation, informed consent forms, data privacy protocols, and safety manuals are available upon request from the corresponding author.

Advanced prediction models.

Building on the basic survival patterns, multivariable prediction models provided more precise risk stratification:

1. - 5-year survival probability for first cousins with hematological disorders: 42.3% (95% CI: 37.8–46.9)
2. - 5-year survival for second cousins with congenital anomalies: 65.1% (95% CI: 60.8–69.3)
3. - Non-consanguineous with sensory impairments: 92.4% (95% CI: 89.7–94.5)

Complete prediction model performance metrics, including calibration and discrimination measures, are provided in S13 Table, and model calibration assessed by comparing predicted versus observed survival probabilities is presented in S9 Fig.

Interaction effects analysis.

Significant interaction was observed between consanguinity and healthcare access (p < 0.001):

1. - High healthcare access: Consanguinity HR = 1.8 (1.4-2.3)
2. - Medium healthcare access: Consanguinity HR = 2.4 (1.9-3.0)
3. - Low healthcare access: Consanguinity HR = 3.2 (2.6-4.0)

Selection bias assessment.

Sensitivity analysis using E-value methodology indicated:

1. - E-value for consanguinity effect: 3.8
2. - Unmeasured confounding would need HR = 3.8 to explain away observed effects
3. - Multiple imputation showed consistent results across missing data patterns

Disease burden analysis.

A comprehensive disease burden assessment revealed substantial impacts attributable to genetic disorders in consanguineous offspring. Disability-Adjusted Life Years (DALYs) were calculated using standard WHO methods, with Years of Life Lost (YLL) based on disorder-specific life expectancies from our cohort and Years Lived with Disability (YLD) applying disability weights from the Global Burden of Disease study. Economic burden estimation incorporated direct medical costs (based on local healthcare pricing), indirect costs (productivity losses using the human capital approach), and out-of-pocket expenditures from household surveys. Sensitivity analyses employed alternative discount rates (1–5%) and disability weights.

The analysis demonstrated:

1. - Total DALYs attributable to consanguinity: 12,450 years
2. - Highest disorder-specific burden: Hematological disorders (4,890 DALYs)
3. - Economic burden estimate: $62.25 million
4. - Cost-effectiveness: Targeted screening showed an ICER of $334 per DALY averted

Detailed disease burden calculations and cost-effectiveness analyses are presented in S14 and S15 Tables.

The cost-effectiveness analysis of targeted screening interventions, including the incremental cost-effectiveness ratio (ICER), is presented in S12 Fig.

Visualization of competing mortality risks.

To visually represent the cause-specific mortality burden, Fig 2 presents the cumulative incidence functions (CIFs) for the three leading causes of death across different age periods. The curves illustrate the probability of death from a specific cause over time, accounting for the competing risk of death from other causes. Notably, congenital anomalies show a steep initial rise in infancy, while hematological disorders become the dominant cause in early childhood (1–5 years). The separation of the curves demonstrates the distinct temporal patterns of mortality for each cause, which would be obscured in a traditional Kaplan-Meier analysis.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Cumulative incidence functions for leading causes of death.

https://doi.org/10.1371/journal.pone.0349764.g002

Cumulative incidence functions for the three leading causes of death among the offspring of consanguineous marriages. The curves show the probability of dying from specific causes while accounting for competing risks from other causes. Congenital anomalies demonstrate rapid onset in infancy, hematological disorders peak in early childhood (1–5 years), and neurological complications show a progressive increase through childhood and adolescence, with age-specific mortality hazard functions visualized in S3 Fig.

Extended results documentation.

Comprehensive results, including complete cohort characteristics, subgroup analyses, sensitivity tests, and extended statistical outputs, are provided in S1–S19 Tables. S1–S13 Figs present detailed survival curves, diagnostic plots, and additional visualizations. Extended analytical results are documented in Supporting Analyses S1–S10 in (S7 File).

Drivers of survival gains and healthcare system interactions.

Our data reveal significant temporal improvements in survival, most dramatically for hemoglobinopathies, where 5-year survival increased from 28.9% to 62.7% across birth cohorts. Interaction analysis indicates that these gains were not uniform; they were most pronounced for disorders amenable to medical management, such as hematological conditions (adjusted HR for the 2013–2024 vs. 1998–2002 cohort = 0.38, 95% CI: 0.29–0.51), compared to structural congenital anomalies where surgical access is limited (HR = 0.62, 95% CI: 0.48–0.80). This pattern underscores that improvements in basic medical care—including more reliable transfusion services, the introduction of iron chelation therapy, and better management of infections—have been the primary drivers of survival gains in our setting. The phased introduction of genetic services, such as prenatal diagnosis post-2015, further reflects this evolving healthcare infrastructure.

A shifting burden and the imperative for integrated care models.

This success heralds a critical shift in public health needs. As more children with severe genetic disorders survive into childhood and beyond, the demand for long-term specialized care, rehabilitation, and psychosocial support will inevitably surge, creating new challenges for healthcare systems [27]. Our findings are part of a broader regional trend observed in the Middle East and North Africa, where national prevention strategies, including mandatory premarital screening programs in countries like Jordan, Egypt, and across the Gulf Cooperation Council states, have successfully reduced the incidence of new affected births [28–33]. This dual reality—declining incidence alongside a growing prevalent cohort of survivors—necessitates a fundamental evolution in health strategy. Public health planning must therefore advance from a primary focus on prevention and reducing infant mortality to a dual, integrated approach. This approach must equally prioritize robust, lifelong care systems, including adult transition clinics and multidisciplinary support services, to address the full spectrum of needs for individuals living with chronic genetic conditions [27].

Phased healthcare framework.

- Primary Prevention (Years 1–2): Community genetic counseling, school-based genetic literacy programs, and engagement of religious leaders for premarital counseling.

- Secondary Prevention (Years 2–3): Expanded newborn screening for hemoglobinopathies, routine prenatal ultrasound, and developmental screening in primary care.

- Tertiary Intervention (Years 3–5): Regional thalassemia centers, specialized pediatric surgical services, and multidisciplinary neurodevelopmental clinics.

Priority action plan.

1. Pre-marital and pre-conception prevention (Immediate): Implement community-based genetic counseling and education programs targeting couples planning marriage, particularly first cousins. Engage religious and community leaders to deliver culturally sensitive messages. Successful models from Qatar and Jordan demonstrate that such approaches can significantly increase awareness and reduce high-risk unions [2,44].
2. Targeted genetic counseling for first-cousin couples (Immediate–Short term): Integrate genetic risk assessment into existing maternal and child health platforms. Provide counseling for at-risk couples identified through family history or community referral.
3. Newborn screening expansion (Short–Medium term): Phase in newborn screening for hemoglobinopathies beginning in high-prevalence districts, leveraging existing Expanded Program on Immunization (EPI) visits for sample collection.
4. Surgical and specialized care (Medium–Long term): Develop regional thalassemia centers and pediatric surgical services for correctable congenital anomalies, particularly cardiac defects that dominate infant mortality.

Economic feasibility in a conflict-affected setting.

The implementation of newborn screening in a resource-limited, conflict-affected setting like Yemen requires careful economic consideration. Our cost-effectiveness analysis (S15 Table) indicates that targeted screening would cost approximately $1,200 per case detected, with an incremental cost-effectiveness ratio (ICER) of $334 per DALY averted—well below the WHO-recommended threshold of one times the GDP per capita (approximately $700 in Yemen). This suggests that even in a fragile health system, phased implementation starting in urban centers with existing laboratory infrastructure is economically feasible and highly cost-effective. A phased approach would allow for gradual scale-up as health system capacity improves.

Strategic research agenda.

Genomic Integration: Whole genome sequencing of cluster representatives, gene-environment interaction studies, and pharmacogenomic profiling for treatment optimization.

Implementation Science: Culturally adapted genetic counseling protocols, integration of genetic services into primary healthcare, and community-based participatory research.

Precision Public Health: Machine learning for risk prediction, mobile health technologies for remote monitoring, and policy modeling for sustainable service delivery.

Global Health Genetics: Comparative studies across consanguineous populations, cross-cultural genetic literacy interventions, and international collaboration for rare variant discovery.

Translational research pipeline.

Short-term (1–2 years): Validation studies, clinical decision support tools, and economic evaluations.

Medium-term (3–5 years): Randomized trials, implementation research, policy development.

Long-term (5 + years): Population impact assessment, sustainable delivery models, global knowledge translation.

Policy recommendations and performance metrics.

1. Integrated Genetic Service Delivery
   1. Tiered counseling system (from community workers to specialists)
   2. Electronic health record flags for high-risk families
   3. Primary care physician training in genetic risk assessment.
2. National Screening Strategy
   
   1. Phase 1: High-prevalence districts (years 1–2)
   2. Phase 2: Regional expansion (years 3–5)
   3. Phase 3: National coverage (years 6+)
   Key performance indicators:
   1. - 40% reduction in consanguinity-associated mortality within 10 years
   2. - 80% coverage of genetic counseling for at-risk couples
   3. - 75% 5-year survival rate for hemoglobinopathies
   4. - 50% reduction in economic burden per DALY

The time for strategic action is now—the health of future generations depends on implementing these evidence-based interventions informed by rigorous longitudinal research. By addressing both the immediate clinical needs and underlying genetic risks, healthcare systems can significantly improve survival and quality of life for affected children while reducing the long-term economic burden on families and communities.