Within- and between-country variation in gestational age thresholds for retinopathy of prematurity screening in Latin America Authors: Juan Carlos Silva, Alejandro Vasquez De Kartzow, Celia Cristina Lomuto, Claudia Zuluaga-Botero, Luz Gordillo, Consuelo Zepeda-Romero, Pedro Acevedo, Veronica Contreras, Guillermo Andrés Monteoliva, Gabriela Saidman, Graham E. Quinn, Clare Gilbert Published: 2026 ABSTRACT Background/objectivesMost Latin American countries have national retinopathy of prematurity (ROP) screening guidelines, but how consistently these thresholds are applied and whether the metric used affects cross-country comparability is poorly documented.Subjects/methodsWe analysed data from the SPROP (Sociedad Panamericana de Retinopatía del Prematuro) database (2021–2024), including 333 of 343 registrants from 17 countries who provided gestational age (GA) and/or birth weight (BW) screening cutoffs. We assessed between- and within-country variation, guideline concordance, and relative variability of GA and BW thresholds using the coefficient of variation (CV). Hospital-level alignment was classified as wider, at, or narrower than the national guideline.ResultsBW cutoffs were 3.6 to 5.1 times more variable than GA cutoffs across all countries. Mean GA cutoffs approximated national guidelines, but with marked country-level differences: Argentina applied broader cutoffs than recommended (Δ = +1.5 weeks; 50.5% permissive), Colombia and Peru applied narrower cutoffs (Δ = −1.9 and − 1.3 weeks; 62.7% and 45.5% restrictive, respectively). Mexico showed a mixed pattern (63.9% at guideline), and Paraguay had the highest adherence (80.0%). Overall deviation from guideline recommendations (permissive plus restrictive) ranged from 20.0% in Paraguay to 66.3% in Argentina.ConclusionsVariation in concordance with national ROP screening guidelines reflects differences in programme implementation as well as national recommendations. Organisational stewardship, coordinated implementation, and transparent institutional policies may support effective screening, while robust local outcome surveillance is essential to guide evidence-based adaptation and future refinement of screening criteria. Introduction Although most Latin American countries have national guidelines for retinopathy of prematurity (ROP) screening, how consistently these guidelines are applied and how much screening thresholds vary within each country is poorly documented. Persistent disparities in ROP-related visual loss between countries highlight the influence of health-system context on outcomes [1]. These disparities have been analysed using the Sociodemographic Index (SDI), a composite measure developed by the Institute for Health Metrics and Evaluation that combines income per capita, educational attainment, and fertility rates in women under 25, underscoring the need for SDI-adapted screening frameworks alongside equitable access to neonatal care, screening, treatment and follow-up [2]. In low- and middle-income settings, more mature infants are at higher risk of severe ROP than in high-income countries due to exposure to modifiable risk factors, underscoring the need for locally informed guidelines [1] and adherence to national screening policies aligned with population-specific risk [3]. Cross- and within-country comparisons depend on the metric used to define screening eligibility. Although both gestational age (GA) and birth weight (BW) are used clinically, GA more directly reflects developmental maturity and retinal risk and can be measured using standardised methods, whereas BW is more sensitive to population-specific growth patterns, maternal health, and intrauterine conditions [4]. We therefore used GA as the primary comparative metric to characterize between- and within-country variation in screening practices and adherence to national guidelines, with BW examined as a secondary metric; it does not model joint GA × BW eligibility or classify deviations by clinical appropriateness. Methods Data source and participants Data were obtained from the SPROP (Sociedad Panamericana de Retinopatía del Prematuro) database (2021–2024), contributed by ophthalmologists and neonatologists responsible for ROP screening in neonatal intensive care units in 17 Latin American countries. Variables included country, institution, profession, and GA and BW screening cutoffs applied at each site. Of 343 registrants, 333 provided complete, usable data; records with missing, inconsistent, or implausible values were excluded, with each registrant contributing a single observation. Brazil was excluded due to insufficient participation for meaningful analysis. Variables and operational definitions The primary variables were the GA cutoff (maximum GA in weeks) and the BW cutoff (maximum weight in grams), which defined screening eligibility at each institution. National ROP screening guidelines were identified through official Ministry of Health or scientific society documents, and each country’s GA threshold served as the benchmark for comparison. Statistical analysis The analysis was conducted in two stages. First, we assessed between- and within-country variation in ROP screening thresholds. Second, we evaluated the relative variability of GA and BW thresholds using the coefficient of variation (CV), calculated as the standard deviation divided by the mean (SD/mean), to identify the most appropriate metric for policy-level comparisons. The CV ratio (CV_BW/CV_GA) was used to compare the dispersion of BW and GA thresholds within each country; values > 1 indicate greater variability in BW thresholds. For countries with at least 20 reporting hospitals (Argentina, Colombia, Mexico, Paraguay, and Peru), screening practices were summarised using the number of units and the range, mean, median, and standard deviation (SD) of GA cutoffs in weeks. The difference between each country’s mean hospital GA cutoff and the national guideline cut-off was expressed as Δ (mean cutoff − national threshold) in weeks. To assess guideline adherence at the hospital level, each hospital’s GA cutoff was classified as wider than the guideline (more permissive), at the guideline, or narrower than the guideline (more restrictive), and the proportion of hospitals in each category was reported per country. Total deviation was defined as the combined percentage of hospitals applying cutoffs either wider or narrower than the national recommendation. Countries with 6–19 reporting hospitals (Chile, Ecuador, Dominican Republic, and Venezuela) were assessed using robust descriptive measures: range, interquartile range (IQR), and median absolute deviation (MAD), given that standard deviation estimates are unreliable in smaller samples. Countries with five or fewer reporting units (Bolivia, Costa Rica, Cuba, El Salvador, Guatemala, Nicaragua, and Panama) were included for completeness and reported descriptively, without formal heterogeneity analyses. All analyses were conducted in RStudio. Scope and limitations This analysis is limited to the variability of applied GA thresholds and their alignment with national guidelines, with a methodological comparison of GA and BW as policy-level metrics. The study does not address combined GA–BW eligibility criteria, the logical rule governing their application (AND vs. OR), programme effectiveness or screening coverage, the clinical appropriateness of deviations from national guidelines, or the association between screening practices and clinical outcomes. These questions are discussed as priorities for future research. Ethics This study involved a retrospective analysis of pre-existing, fully de-identified data containing no information that could reasonably be used to identify individual participants. In accordance with applicable regulations and institutional policies governing minimal-risk research, formal ethics committee review was waived. Generative AI was used solely to improve the manuscript’s clarity and readability. All scientific content, analyses, and conclusions were developed, verified, and approved by the authors. Results After exclusions, 324 registrants were included. Of these, 268 were from countries with 20 or more reporting hospitals, 35 from countries with 6–19 hospitals, and 21 from countries with five or fewer. Variability of gestational age and birthweight thresholds across countries Birth weight thresholds were substantially more variable than GA thresholds in every country (Table 1). CV values for GA ranged from 0.03 (Paraguay) to 0.07 (Argentina), while CV values for BW ranged from 0.13 (Paraguay) to 0.35 (Argentina). The CV ratio (CV_BW/CV_GA) exceeded 1.0 in all countries, ranging from 3.6 (Colombia) to 5.1 (Argentina), indicating that BW thresholds varied 3.6 to 5.1 times more across hospitals than GA thresholds. Given the lower between-country variability in GA screening thresholds compared with BW thresholds (Table 1), GA was used in Table 2 as the more consistent parameter for comparing adherence to national screening guidelines across countries; this does not diminish the importance of BW as a criterion for screening eligibility and ROP risk assessment. Table 1Relative variability of gestational age and birth weight screening thresholds between countriesCountry n Mean GA(weeks)SD GA(weeks)CV GAMean BW(g)SD BW(g)CV BWCV ratio(BW/GA)Argentina10133,52,30,071 737.6608,70,355,14Colombia6734,11,90,061 947.8388,10,23,6Mexico36341,60,051 805.6333,30,183,96Paraguay20340,90,031 825.0230,90,134,69Peru4434,71,80,051 977.3384,40,193,8The CV ratio is partly influenced by the nonlinear relationship between fetal weight and gestational age and should be interpreted alongside the absolute CV values for each metricNational ROP screening criteria for gestational age (GA) and birth weight (BW) were: Argentina, GA ≤ 32 weeks/BW < 1,500 g; Colombia, GA < 36 weeks/BW < 2,000 g; Mexico, GA ≤ 34 weeks/BW < 1,750 g; Paraguay, GA ≤ 34 weeks/BW ≤ 1,750 g; and Peru, GA < 36 weeks/BW ≤ 2,000 gAbbreviations: GA = gestational age (weeks); BW = birth weight (grams); SD = standard deviation; CV = coefficient of variation; n = number of neonatal units per country Table 1Relative variability of gestational age and birth weight screening thresholds between countriesCountry n Mean GA(weeks)SD GA(weeks)CV GAMean BW(g)SD BW(g)CV BWCV ratio(BW/GA)Argentina10133,52,30,071 737.6608,70,355,14Colombia6734,11,90,061 947.8388,10,23,6Mexico36341,60,051 805.6333,30,183,96Paraguay20340,90,031 825.0230,90,134,69Peru4434,71,80,051 977.3384,40,193,8The CV ratio is partly influenced by the nonlinear relationship between fetal weight and gestational age and should be interpreted alongside the absolute CV values for each metricNational ROP screening criteria for gestational age (GA) and birth weight (BW) were: Argentina, GA ≤ 32 weeks/BW < 1,500 g; Colombia, GA < 36 weeks/BW < 2,000 g; Mexico, GA ≤ 34 weeks/BW < 1,750 g; Paraguay, GA ≤ 34 weeks/BW ≤ 1,750 g; and Peru, GA < 36 weeks/BW ≤ 2,000 gAbbreviations: GA = gestational age (weeks); BW = birth weight (grams); SD = standard deviation; CV = coefficient of variation; n = number of neonatal units per country Relative variability of gestational age and birth weight screening thresholds between countries The CV ratio is partly influenced by the nonlinear relationship between fetal weight and gestational age and should be interpreted alongside the absolute CV values for each metric National ROP screening criteria for gestational age (GA) and birth weight (BW) were: Argentina, GA ≤ 32 weeks/BW < 1,500 g; Colombia, GA < 36 weeks/BW < 2,000 g; Mexico, GA ≤ 34 weeks/BW < 1,750 g; Paraguay, GA ≤ 34 weeks/BW ≤ 1,750 g; and Peru, GA < 36 weeks/BW ≤ 2,000 g Abbreviations: GA = gestational age (weeks); BW = birth weight (grams); SD = standard deviation; CV = coefficient of variation; n = number of neonatal units per country Table 2Screening gestational age cutoffs used in neonatal units in five countries relative to the national screening guidelinesCountry n Mean GA cutoff(weeks)SD(weeks)Range(weeks)National guideline(weeks)Δ from guideline(weeks)Argentina10133,52,330–36≤32+1.5Colombia6734,11,930–38≤36−1.9Mexico36341,630–36≤340Paraguay20340,932–36≤340Peru4434,71,830–36≤36−1.3Δ = difference between the mean GA cutoff used by neonatal units and the national guideline threshold (weeks). Positive values indicate more permissive practice than the guideline; negative values indicate stricter practice. A Δ of 0 indicates alignment of the mean with the national guideline but does not preclude variability among individual hospitals (see Table 3 for hospital-level alignment)Abbreviations: GA = gestational age; SD = standard deviation; n = number of neonatal units per country Table 2Screening gestational age cutoffs used in neonatal units in five countries relative to the national screening guidelinesCountry n Mean GA cutoff(weeks)SD(weeks)Range(weeks)National guideline(weeks)Δ from guideline(weeks)Argentina10133,52,330–36≤32+1.5Colombia6734,11,930–38≤36−1.9Mexico36341,630–36≤340Paraguay20340,932–36≤340Peru4434,71,830–36≤36−1.3Δ = difference between the mean GA cutoff used by neonatal units and the national guideline threshold (weeks). Positive values indicate more permissive practice than the guideline; negative values indicate stricter practice. A Δ of 0 indicates alignment of the mean with the national guideline but does not preclude variability among individual hospitals (see Table 3 for hospital-level alignment)Abbreviations: GA = gestational age; SD = standard deviation; n = number of neonatal units per country Screening gestational age cutoffs used in neonatal units in five countries relative to the national screening guidelines Δ = difference between the mean GA cutoff used by neonatal units and the national guideline threshold (weeks). Positive values indicate more permissive practice than the guideline; negative values indicate stricter practice. A Δ of 0 indicates alignment of the mean with the national guideline but does not preclude variability among individual hospitals (see Table 3 for hospital-level alignment) Abbreviations: GA = gestational age; SD = standard deviation; n = number of neonatal units per country Mean gestational age cutoffs and deviation from national guidelines Mean GA cutoffs in the five countries with the most participants approximated national guidelines, yet within-country variation was substantial (Table 2). Argentina showed the largest deviation, with a mean cutoff 1.5 weeks above the national threshold (33.5 vs. ≤32 weeks). Colombia and Peru applied cutoffs below their national thresholds (Δ = −1.9 and − 1.3 weeks, respectively), while Mexico and Paraguay showed mean cutoffs aligned with their national guidelines (Δ = 0). Within-country ranges of GA cutoffs ranged from four weeks in Paraguay (32–36 weeks) to eight weeks in Colombia (30–38 weeks), indicating substantial hospital-level heterogeneity regardless of mean alignment with national guidelines. Hospital-level alignment with national guidelines Hospital-level adherence to national guidelines varied considerably across countries (Table 3). In Argentina, half of the hospitals (50.5%) applied GA cutoffs wider than the national guideline, 33.7% followed the recommended threshold, and 15.8% used stricter cutoffs. Colombia showed the opposite pattern, with 62.7% of hospitals applying narrower cutoffs than the guideline and only 1.5% applying wider cutoffs. Peru similarly favoured stricter practice, with 45.5% of hospitals using narrower cutoffs and none using wider cutoffs. Mexico and Paraguay showed the highest guideline adherence, with 63.9% and 80.0% of hospitals at the recommended threshold, respectively, though minor bidirectional deviations were present in both countries. Table 3Hospital-level alignment of gestational age screening cutoffs with national guidelines in five countriesCountry n Wider than guideline(more permissive)n (%)At guidelinen (%)Narrower than guideline(more restrictive)n (%)Argentina10151 (50.5%)34 (33.7%)16 (15.8%)Colombia671 (1.5%)24 (35.8%)42 (62.7%)Mexico368 (22.2%)23 (63.9%)5 (13.9%)Paraguay202 (10.0%)16 (80.0%)2 (10.0%)Peru440 (0.0%)24 (54.5%)20 (45.5%)Values represent the number of neonatal units and percentage within each country. Wider than guideline = GA cutoff above the national threshold (more permissive); narrower than guideline = GA cutoff below the national threshold (more restrictive). Total deviation per country = sum of wider and narrower hospitals as a percentage of all reporting unitsAbbreviations: GA = gestational age; n = number of neonatal units per country Table 3Hospital-level alignment of gestational age screening cutoffs with national guidelines in five countriesCountry n Wider than guideline(more permissive)n (%)At guidelinen (%)Narrower than guideline(more restrictive)n (%)Argentina10151 (50.5%)34 (33.7%)16 (15.8%)Colombia671 (1.5%)24 (35.8%)42 (62.7%)Mexico368 (22.2%)23 (63.9%)5 (13.9%)Paraguay202 (10.0%)16 (80.0%)2 (10.0%)Peru440 (0.0%)24 (54.5%)20 (45.5%)Values represent the number of neonatal units and percentage within each country. Wider than guideline = GA cutoff above the national threshold (more permissive); narrower than guideline = GA cutoff below the national threshold (more restrictive). Total deviation per country = sum of wider and narrower hospitals as a percentage of all reporting unitsAbbreviations: GA = gestational age; n = number of neonatal units per country Hospital-level alignment of gestational age screening cutoffs with national guidelines in five countries Values represent the number of neonatal units and percentage within each country. Wider than guideline = GA cutoff above the national threshold (more permissive); narrower than guideline = GA cutoff below the national threshold (more restrictive). Total deviation per country = sum of wider and narrower hospitals as a percentage of all reporting units Abbreviations: GA = gestational age; n = number of neonatal units per country Patterns of deviation across countries The total proportion of hospitals deviating from the national guideline in either direction ranged from 20.0% in Paraguay to 66.3% in Argentina, with intermediate values in Mexico (36.1%), Peru (45.5%), and Colombia (64.2%) (Table 3). Deviation was predominantly in the permissive direction in Argentina, predominantly restrictive in Colombia and Peru, bidirectional in Mexico, and minimal in both directions in Paraguay. Within-country variability among smaller samples In countries with 6–19 reporting hospitals, GA cutoffs were assessed descriptively. Chile (n = 8) showed complete uniformity, with all hospitals applying a cutoff of 32 weeks. Ecuador (n = 9) had the widest spread, with cutoffs ranging from 30 to 38 weeks. The Dominican Republic and Venezuela (n = 9 each) showed limited variation, with cutoffs clustered between 34 and 36 weeks. Countries with five or fewer reporting units are reported for completeness only. Discussion Our findings support using GA rather than BW as the primary metric for assessing adherence to ROP screening policies in hospitals and countries. BW thresholds varied 3.6 to 5.1 times more than GA thresholds in every country included. Because BW is shaped by fetal growth, maternal health, and local socioeconomic conditions, the same BW does not represent the same level of risk of ROP in different populations. GA, by contrast, directly reflects developmental maturity and is biologically comparable across settings. Our findings should not be interpreted as indicating that GA is a better predictor of ROP risk or severity than BW. Birth weight and postnatal growth remain important predictors of ROP outcomes and are relevant to screening eligibility and individual risk assessment. Rather, our findings suggest that GA may provide a more consistent metric for assessing adherence to existing screening protocols across settings, given the greater variability observed in BW-based screening criteria. In Latin America, mean GA cutoffs generally approximated national guidelines, reflecting broad agreement on the biological window of risk for severe ROP. However, within-country GA ranges spanned four weeks in Paraguay and eight weeks in Colombia, revealing substantial hospital-level heterogeneity even among clinicians operating under the same national policy. Marked variability in ROP risk between neonatal intensive care units, even within the same city, has been well documented in the region [5]. Rapid expansion of neonatal services, fragmented care delivery, and resource constraints have driven variability in the quality of care, including oxygen management, contributing to heterogeneous ROP risk between neonatal units [6]. Similar heterogeneity has been reported in Asia [7] and Europe [8], supporting the view that universal screening criteria are neither feasible nor desirable [9] due to local differences in case mix, neonatal survival and the quality of care justify context-driven adaptation. Non-compliance with national screening cut-offs should not automatically be interpreted as poor adherence or lack of knowledge. In our study, some hospitals, particularly where national guidelines are relatively restrictive, broader screening criteria may represent deliberate, evidence-informed local adaptations based on neonatal outcomes and service-specific experience; Argentina exemplifies this pattern, with many hospitals applying broader GA criteria, possibly reflecting incorporation of additional risk factors within a highly organized screening program. Conversely, where national guidelines are more permissive, individual services with robust local outcome data may appropriately adopt more restrictive criteria to optimise resource use without compromising patient safety, as illustrated by Colombia and Peru, while Mexico suggests that national guideline development combined with broad public insurance coverage through IMSS may facilitate consistent practice. Both Ecuador and Paraguay had small numbers of participating hospitals; however, their concordance rates differed markedly. In Paraguay, consensus between neonatologists and ophthalmologists within a coordinated national programme may account for high guideline concordance. In Ecuador, a small number of participating hospitals alone was insufficient to ensure consistency, underscoring the importance of structured programme coordination and governance. Together, these findings suggest that variation in screening practice should be interpreted within the context of local evidence, institutional stewardship, health system organisation, and governance rather than viewed uniformly as poor compliance. When deviations from national guidelines occur, they are best supported by robust local outcome data, multidisciplinary review, documented institutional approval, and ongoing monitoring of clinical outcomes. From a medico-legal perspective, departures from national recommendations are more readily justified when they reflect transparent, evidence-based institutional policies with documented rationale and outcome evaluation. In high-income settings, screening criteria are increasingly adapted to local contexts to reduce unnecessary examinations without compromising the detection of treatable disease [10, 11]. Canada has retained conservative safeguards within risk-informed approaches even in well-resourced settings [12–14], while Sweden narrowed eligibility only after consolidating national registries, stable outcomes, and real-time surveillance [15]. Both examples underscore that future refinement of screening criteria should be grounded in robust local outcome data, particularly on the incidence and characteristics of treatment-requiring ROP. This study makes several contributions to the implementation of ROP screening. First, it provides the first multicountry assessment of concordance between reported hospital screening practices and national guidelines in Latin America using the SPROP registry. Second, it demonstrates that concordance is influenced not only by national recommendations but also by institutional stewardship, program coordination, health system organization, and governance. Third, by distinguishing permissive from restrictive deviations, it offers a framework for interpreting non-compliance based on local evidence rather than assuming poor adherence. Fourth, it suggests that in regions with substantial variation in neonatal care capacity, complete guideline concordance should not be the sole indicator of programme quality, as evidence-based local adaptation may improve both safety and efficiency. Taken together, these findings position SPROP registry data as a practical tool for ongoing policy monitoring and highlight the need for country-specific implementation strategies and robust local outcome data to strengthen adherence to and future refinement of national ROP screening guidelines. However, the study was not designed to determine whether the observed variability reflects transparent, evidence-based institutional policies, insufficient methodological rigour, or non-compliance with national guidelines. A further limitation is that the findings are based solely on SPROP-affiliated centres and therefore may not fully represent all ROP services within the participating countries. Although GA may provide a more consistent basis than BW for comparing adherence to ROP screening criteria across countries, its reliability depends on accurate gestational dating. Access to prenatal care, availability and timing of obstetric ultrasound, and the quality of maternal and neonatal records may vary both between and within countries, potentially resulting in misclassification of GA. This limitation should be considered when using GA as a standardized indicator for comparing adherence to ROP screening protocols across settings. Future studies linking facility-specific screening thresholds and adherence to national guidelines with patient-level ROP outcomes are needed to determine whether variability in GA and BW screening criteria translates into differences in ROP detection, progression, treatment-requiring disease, or other clinically relevant outcomes. In summary, national and international ROP guidelines adopt broader eligibility criteria to prioritise safety and equity in settings where data may underrepresent the true population at risk [16]. As data quality, surveillance capacity, and health-system performance improve, responsible evolution of screening criteria becomes possible. At the service level, any adaptation should be grounded in complete local data including GA distributions, screening coverage, characteristics of infants developing treatment-requiring ROP, and timing of detection and treatment to ensure that narrowing eligibility does not compromise timely identification of infants who need treatment.