Abstract
The red reflex is an integral component of routine newborn screening in high-income countries. It is a non-invasive screening test using an ophthalmoscope to check for a normal red or orange glow from a newborn s pupils. However, this is yet to be adopted as a routine examination in many low- and middle- income countries. Pursuant to the asymptomatic nature of many congenital and acquired ocular diseases, it is imperative to have contextual data to facilitate upscaling of the use of this simple examination method for identification of ocular pathologies.
To determine the proportion and causes of abnormal red reflex in neonates born at Moi Teaching and Referral Hospital (MTRH), and the accuracy of the red reflex in identifying these neonates.
This was a cross-sectional hospital-based study of 552 consecutively sampled neonates between February 2024 and September 2025 at the newborn unit of Moi Teaching and Referral Hospital in western Kenya. Initial screening of the red reflex was done using a direct ophthalmoscope. The reference test was dilatation and indirect ophthalmoscopy (gold standard) done by a specialist ophthalmologist. The red reflex was classified as normal or white or abnormal with abnormal defined as dull or absent. Sensitivity, specificity, positive and negative predictive values determined the accuracy of this reflex.
A total of 552 neonates were enrolled; 45.8 % female and 54.2% were male. About two thirds (66.8%) of the neonates had a birthweight >2.5kgs and 22.3% weighed lower than 1.5kgs. More than half (56%) were borne at ≥ 39 week gestation, while a 5.6% were born at <28 weeks. The red reflex was performed on 1104 eyes of the 552 enrolled neonates. An abnormal red reflex was identified in 124 neonates (22.5%) of whom 95(76.6%) had a confirmed abnormality on ophthalmoscopy. The most common identified causes were retinal haemorrhage (11.9%), physiologic retinal immaturity (9.2%) and retinopathy of prematurity (4.0%). Overall diagnostic accuracy of the red reflex examination was 83.7% with a sensitivity of 60.9%, specificity of 92.7%, positive predictive value of 76.6%, and negative predictive value of 85.7%. Gestational age (<28 weeks, aOR 4.1(95% CI,1.4-12.5), and birth weight (<1.5kgs, aOR 6.29(95% CI,3.0-14.5)) were significantly associated with an abnormal red reflex.
The red reflex has high accuracy and specificity and good sensitivity. The findings of this study affirm the utility of the red reflex in identifying neonates with ocular pathologies in a low resource setting. Integrating this test as a screening tool for comprehensive eye care will lead to prompt referral for specialist care, a reduction in blindness from neonatal ocular pathologies and improved outcomes.
Author Contributions
Copyright© 2026
Njie Albertine Enjema, et al.
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Competing interests The authors declare no conflict of interest.
Funding Interests:
Citation:
Introduction
Newborn screening has steadily become an integral part of fostering neonatal care in the developed world. However, many low- and middle- income countries are yet to incorporate the full range of newborn screening tests when compared with higher income countries that are able to screen for about fifty diseases at birth Globally, childhood blindness has been shown to have a linear relationship with under-five mortalities The red reflex, an examination of pupil reflections, can help identify avoidable causes of blindness such as congenital cataract, retinoblastoma, retinal detachment, and vitreous opacities Visual impairment has a huge impact on a child’s academic and career opportunities, social life including stigmatization, functionality, and overall health
Materials And Methods
This was a cross-sectional hospital-based single-blinded study of 552 neonates in the newborn unit of the Riley Mother and Baby section of Moi Teaching and Referral Hospital. On average 150 babies are managed at the newborn unit every month. The newborn unit comprises of the following departments: a neonatal intensive care, acute prematurity, stable prematurity, stable term, high-dependency care and surgical. It is managed by 4 neonatologists, 2 consultant pediatricians and a variable number of neonatology fellows, pediatric residents, medical officers, medical officer interns, nurses, nutritionist and clinical officers. We enrolled 552 neonates consecutively between February 2024 to September 2025. Sample size estimation was done using the Burderer s formula imputing : 23% as a sensitivity, 98% as specificity, Ethical approval from Institutional Review and Ethics Committee (IREC) of Moi University and Moi Teaching and Referral Hospital (approval number: 0004553) and administrative approval from the CEO of MTRH were obtained. Codes were assigned to the enrolled neonates for utilization in this study and for enhancement of confidentiality. Written informed consent was obtained from parents or guardians of neonates for ocular examinations to be done. Consent was done in English or Kiswahili as preferred by parent/guardian of neonates. Assent was not applicable in this study. Prior to use, the consent form was translated from English to Kiswahili and back translated from Kiswahili to English for ease of understanding by parents/guardians of neonates. Term neonates and pre-term neonates who had attained at least one month of age were enrolled in this study. Enrolment was done at one month of age for the pre-term neonates to eliminate the bias of high false negative screening in retinopathy of prematurity at one month. Six neonates were not enrolled as a result of refusal of consent and three neonates were excluded due to anophthalmia. All neonates enrolled in this study underwent red reflex assessment using a direct ophthalmoscope by a research assistant in a dark room at time of discharge from the hospital or after one month of age for preterm neonates admitted to the newborn unit. The research assistant was a qualified registered degree level nurse who underwent training by a certified ophthalmologist on conducting of the red reflex. The color, brightness and equality of the reflex in both eyes of each participant was observed and documented. The reflex results were subsequently categorized as normal, white or abnormal (dull or absent )reflex. The outcome of red reflex was categorized as dichotomous normal or abnormal (dull or absent) for inclusion in the 2x2 table for estimation of accuracy. The standard operating procedure for conduction of the red reflex was adopted from Volume 27 of the Community Eye Health Journal guidelines on how to test for the red reflex in a child (Community Eye Health Journal, 2014). The red reflex by direct ophthalmoscope was considered as the primary test. Additionally, the Hawthorne effect on the user of the initial test was minimized by the fact that all neonates were screened (were asymptomatic) and thus, there were no ocular symptoms or signs to influence the initial test user on if the red reflex should be reported as normal or absent/abnormal on the basis of clinical presentation. The confirmatory test was dilatation and indirect ophthalmoscopy (reference test) by an experienced specialist ophthalmologist using an ophthalmoscope. All neonates, irrespective of initial test (red reflex assessment) results, underwent the reference standard test (dilatation and indirect ophthalmoscopy). Dilatation was done using Tropicamide-phenylephrine a maximum of two drops was instilled in each eye, with a 5-minute interval between drops, resulting in adequate pupil dilation within 15-30 minutes, which persisted for up to 3 hours. During the eye examination, proparacaine eye drops were administered to provide topical anaesthesia, thereby reducing pain and discomfort and improving tolerance of the procedure in neonates. Following completion of the eye examination, moxifloxacin eye drops were administered as prophylaxis. Moxifloxacin eye drops were used as an infection prevention method given that the study was conducted on neonates including premature neonates, who consist of a vulnerable population. In addition, the speculums used were not disposable speculums and had to be sterilized before utilisation on each neonate. Sterilization however does not guarantee 100% infection prevention and thus the need for ocular antibiotic prophylaxis. Futhermore, moxifloxacin is abroad spectrum antibiotic and thus was a great choice for prophylaxis in the current study. The identification of an ocular pathology was also documented as dichotomous (present or absent) to complete the 2X2 contingency table for assessment of diagnostic accuracy of the red reflex test. This was a single-blinded study, as the ophthalmologists carrying out the reference test were unaware of the results of the primary test. In addition, the study participants consisted of a non -enriched sample due to the lack of heterogeneity in assessment of the neonates, i.e., only one index test and one reference standard test was done. All study data were captured using Redcap, a secure web-based application for managing research data. Regular quality checks were conducted throughout the data collection period to ensure completeness, accuracy, and consistency of the data. Following these checks, a clean Comma-Separated Values (CSV) file was exported from Redcap for subsequent analysis. Descriptive statistics were generated to summarize neonates characteristics. Categorical variables were presented as frequencies and proportions with corresponding 95% confidence intervals estimated using binomial exact methods. Continuous variables were summarized using means and standard deviations or medians and interquartile ranges (IQR), depending on variable distribution. Group comparisons were conducted using Pearson s Chi-square test for categorical variables and analysis of variance (ANOVA) for continuous variables, with simulation-based p-value correction applied when expected cell counts were small. The primary outcome for inferential analysis was the presence of abnormal red reflex (binary: abnormal vs. normal). To assess associations between neonatal and clinical characteristics and the primary outcome, univariable and multivariable logistic regression models were fitted. Crude odds ratios (ORs) and adjusted (aORs) with 95% confidence intervals were reported. All variable in consideration that were deemed to be clinically relevant were included in the multivariable logistic regression model to estimate aORs. Model diagnostics were performed to assess overall model adequacy. Diagnostic accuracy of the initial red reflex test was evaluated using the ophthalmologist-confirmed assessment as the reference standard. A 2×2 confusion matrix was constructed to compare results from the initial screening with those from the confirmatory ophthalmologic examination. The analysis of accuracy of the red reflex in comparison to the reference test was done at neonate level and not at individual eye level. Therefore, the results from the two eyes of each neonate were considered together, and each neonate contributed one observation to the comparison between the red reflex test and the reference standard. Sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and overall accuracy were calculated along with their corresponding 95% confidence intervals. Sensitivity was defined as the proportion of true abnormal cases correctly identified by the screening test (TP/(TP+FN)), while specificity represented the proportion of true normal cases correctly identified (TN/(TN+FP)). Positive and negative predictive values were calculated as the probability that positive and negative screening outcomes, respectively, reflected true findings based on the reference standard. Overall diagnostic accuracy was defined as the proportion of correctly classified observations (TP+TN) out of the total screened population. Wilson score confidence intervals were applied to all diagnostic accuracy metrics to account for binomial distribution characteristics and improve interval coverage over the Wald method, particularly for proportions near 0 or 1 or small cell counts. All analyses were conducted using the caret and binom statistical packages in R version 4.4.1
Results
We recruited 552 neonates between February 2024 to September 2025. The red reflex was assessed in 1104 eyes of the 552 enrolled neonates using a direct ophthalmoscope, with slightly more than half of the population being male(54.2%)and females being(45.8%). About two thirds (66.8%) of the neonates had a birthweight >2.5kgs and 22.3% weighed lower than 1.5kgs. More than half (56%) were born at ≥ 39 week’ gestation, while a 5.6% were born at <28 weeks. An abnormal red reflex was identified in 124 neonates, resulting in a prevalence of 22.5% (95% Confidence Interval: 19.0% to 26.2%). Consequently, 428 neonates (77.6%, 95% CI: 73.8% to 81.0%) presented with a normal red reflex. Bivariate analysis ( Using multivariable logistic regression analysis ( Following confirmatory examination with dilatation and indirect ophthalmoscopy (the reference standard), several ocular pathologies were identified ( The diagnostic performance of the initial red reflex test, using ophthalmologist-confirmed findings as the reference standard, is presented in Yes means pathology identified and no refers to normal ocular findings Based on these results, the diagnostic accuracy of the Red reflex test was as follows: sensitivity 60.9% (95% CI: 53.1–68.2), specificity 92.7% (95% CI: 89.7–94.9), positive predictive value76.6% (95% CI: 68.4–83.2), negative predictive value 85.7% (95% CI: 82.1–88.7), and overall accuracy 83.7% (95% CI: 80.4–86.5) ( In a subsample of 211 neonates (100 normal, 111 abnormal by reference test), the neonate’s history and relevant covariates were examined for their association with confirmed ocular disorders ( Among the 111 neonates with a confirmed ocular disorder who were followed up, the majority (n=86, 77.5%) had a normal outcome at their first follow-up visit, indicating resolution or regression of the initial finding. However, significant adverse outcomes were also noted: three neonates (2.7%) died, two (1.8%) developed bilateral blindness secondary to advanced retinopathy of prematurity, and nine (8.1%) had ongoing ROP requiring monitoring. An additional 10 neonates (9.0%) missed their scheduled follow-up appointments. Assessing differences in other diagnoses between the normal vs non normal neonates based on reference test. Neonates with confirmed ocular pathology most frequently carried secondary diagnoses of respiratory distress syndrome (RDS), Prematurity, Neonatal sepsis (NNS), Transient Tachypnea of the Newborn (TTN) and Extremely Low Birth Weight (ELBW). Notable comorbidities also included Hypoxic-Ischemic Encephalopathy (HIE), very low birth weight (VLBW), and neonatal seizures. In contrast, neonates without ocular pathology exhibited a similar, broad range of secondary diagnoses, prominently featuring Prematurity, RDS, and NNS/NNU-related care, along with diagnoses like Meconium Aspiration Syndrome (MAS), congenital anomalies (e.g., Spina Bifida), and congenital heart disease (CHD). This overlap underscores that both groups were drawn from a high-risk neonatal population with substantial comorbidity burdens.
Abnormal
Normal
Total (N=552)
p value
<28
24 (77.4%)
7 (22.6%)
31 (100.0%)
< 0.001
28-34
30 (31.2%)
66 (68.8%)
96 (100.0%)
35-38
16 (11.9%)
119 (88.1%)
135 (100.0%)
39+
54 (18.6%)
236 (81.4%)
290 (100.0%)
Male
68 (22.7%)
231 (77.3%)
299 (100.0%)
0.916
Female
56 (22.1%)
197 (77.9%)
253 (100.0%)
1-7 days
39 (15.5%)
213 (84.5%)
252 (100.0%)
< 0.001
After 7 days
85 (28.3%)
215 (71.7%)
300 (100.0%)
<1.5kg
63 (51.2%)
60 (48.8%)
123 (100.0%)
< 0.001
1.5-2.5kgs
6 (10.0%)
54 (90.0%)
60 (100.0%)
>2.5kgs
55 (14.9%)
314 (85.1%)
369 (100.0%)
Abnormal
95 (60.9%)
61 (39.1%)
156 (100.0%)
< 0.001
Normal
29 (7.3%)
367 (92.7%)
396 (100.0%)
Dependent:
OR (univariable)
OR (multivariable)
Gestational Age (LNMP)
39+
-
-
35-38
0.59 (0.31-1.05, p=0.082)
0.70 (0.37-1.29, p=0.273)
28-34
1.99 (1.17-3.34, p=0.010)
0.74 (0.33-1.64, p=0.466)
<28
14.98 (6.44-39.32, p<0.001)
4.01 (1.37-12.54, p=0.013)
Sex of the baby
Male
-
-
Female
0.97 (0.64-1.44, p=0.865)
0.65 (0.41-1.04, p=0.075)
Day of life
1-7 days
-
-
After 7 days
2.16 (1.42-3.33, p<0.001)
0.79 (0.43-1.40, p=0.427)
Birth weight (g)
>2.5kgs
-
-
1.5-2.5kgs
0.63 (0.24-1.44, p=0.317)
0.86 (0.30-2.11, p=0.759)
<1.5kg
5.99 (3.82-9.49, p<0.001)
6.29 (2.80-14.54, p<0.001)
Ocular Condition
Count (%)
Retina haemorrhages
66 (11.9%)
Retinopathy of prematurity
22 (4.0%)
Failed dilatation of unspecified etiology
12 (2.2%)
Vitreous opacity
6 (1.1%)
Persistent fetal vasculature
5 (0.9%)
Engorged vessels
3 (0.5%)
Pale atrophic disc
3 (0.5%)
Large macula breed
2 (0.4%)
Tigroid fundus
2 (0.4%)
Congenital cataract
1 (0.2%)
Retinal detachment
1 (0.2%)
Disc pallor
1 (0.2%)
Radial posterior synechiae
1 (0.2%)
Perforated central cornea with fibrinous exudate
1 (0.2%)
Immature retina (physiologic)
51 (9.2%)
Red reflex findings
Reference test: Yes
Reference test: No
Yes
95 (76.6%)
29 (23.4%)
No
61 (14.3%)
367 (85.7%)
Total
156
396
Measure
CI
Sensitivity
60.9% (53.1–68.2)
Specificity
92.7% (89.7–94.9)
Positive Predictive Value
76.6% (68.4–83.2)
Negative Predictive Value
85.7% (82.1–88.7)
Accuracy
83.7% (80.4–86.5)
Abnormal
Normal
Total
p value
AVD
14 (87.5%)
2 (12.5%)
16 (100.0%)
< 0.001
C/S
29 (36.7%)
50 (63.3%)
79 (100.0%)
SVD
68 (58.6%)
48 (41.4%)
116 (100.0%)
0
48 (43.2%)
15 (15.0%)
63 (29.9%)
0.135
1
33 (29.7%)
48 (48.0%)
81 (38.4%)
2
16 (14.4%)
22 (22.0%)
38 (18.0%)
3
7 (6.3%)
12 (12.0%)
19 (9.0%)
4+
7 (6.3%)
3 (3.0%)
10 (4.7%)
E-E-E
63 (50.4%)
62 (49.6%)
125 (100.0%)
0.216
E-E-M
1 (50.0%)
1 (50.0%)
2 (100.0%)
M-E-E
21 (67.7%)
10 (32.3%)
31 (100.0%)
M-E-M
1 (100.0%)
0 (0.0%)
1 (100.0%)
M-M-E
9 (56.2%)
7 (43.8%)
16 (100.0%)
M-M-M
10 (55.6%)
8 (44.4%)
18 (100.0%)
S-E-E
1 (50.0%)
1 (50.0%)
2 (100.0%)
S-M-E
1 (50.0%)
1 (50.0%)
2 (100.0%)
S-M-M
2 (100.0%)
0 (0.0%)
2 (100.0%)
S-S-M
0 (0.0%)
1 (100.0%)
1 (100.0%)
NA
2 (18.2%)
9 (81.8%)
11 (100.0%)
No
27 (36.0%)
48 (64.0%)
75 (100.0%)
< 0.001
Yes
84 (61.8%)
52 (38.2%)
136 (100.0%)
Awaits Cataract Surgery
1 (100.0%)
0 (0.0%)
1 (100.0%)
Bilateral blindness secondary to ROP stage 5-
1 (100.0%)
0 (0.0%)
1 (100.0%)
Died
3 (100.0%)
0 (0.0%)
3 (100.0%)
Missed appointment
10 (100.0%)
0 (0.0%)
10 (100.0%)
Normal
86 (100.0%)
0 (0.0%)
86 (100.0%)
Ongoing ROP regress on the R.E, LE-Blindness
1 (100.0%)
0 (0.0%)
1 (100.0%)
Ongoing ROP regression
6 (100.0%)
0 (0.0%)
6 (100.0%)
Ongoing ROP regression, later baby died
1 (100.0%)
0 (0.0%)
1 (100.0%)
Ongoing ROP regression.
1 (100.0%)
0 (0.0%)
1 (100.0%)
R.E-Normal. L.E -Blindness
1 (100.0%)
0 (0.0%)
1 (100.0%)
Discussion
This study was conducted to determine the proportion, causes, and diagnostic accuracy of the red reflex test among neonates at MTRH. The proportion of neonates with abnormal red reflex was 22.5%, and 28.3% of neonates were confirmed to have ocular pathology on reference examination. The most common abnormalities were retinal hemorrhage (11.9%) and immature retina (9.2%). Extreme prematurity (<28 weeks of gestation) and very low birth weight (<1.5 kg) were strong independent predictors of abnormal reflex. The red reflex test demonstrated moderate sensitivity (76.6%), high specificity (85.7%), and an excellent negative predictive value (92.7%). These metrics confirmed its value as a screening tool in this setting, consistent with existing studies In our study, 22.5% of neonates had an abnormal red reflex. Other studies have reported lower proportions between 0.5% and 5%. Congenital cataract and retinoblastoma were rare in our study, but their detection is critical. Given that these conditions can hamper the quality of life and have negative outcomes, even isolated cases of these conditions carry lifelong consequences, and their rarity highlight the need for routine newborn ocular screening. Other studies have reported higher detection of congenital cataract beyond the neonatal period such as by Singh et al who reported 77.9% of their study population with bilateral cataract at a mean age of 4.1 months [±2.6 months] Evidence-based data on the diagnosis of ocular pathologies emphasize the fact that rare but severe causes of leukocoria are often missed without structured screening programs Extreme prematurity and very low birth weight were the strongest independent predictors of an abnormal red reflex. Neonates born before 28 weeks of gestation were four times more likely to have an abnormal reflex, while those weighing less than 1.5 kg had six times the odds. These findings can be explained by the fact that this group of neonates are more likely to have an underdeveloped retinal vasculature. In addition, an underdeveloped retina poses a higher risk of development of retinopathy of prematurity Gender and day of life at examination were not significant predictors of an abnormal red reflex in the multivariable analyses. This may probably suggest that intrinsic neonatal maturity outweighed demographic or temporal factors. Similar findings were observed by Blazon et al., and Chen et al., who reported that sex and age at screening were not significant factors that affected identification of ocular pathologies by the red reflex test. The red reflex test had moderate sensitivity and high specificity. Sensitivity of 60.9% indicated that approximately one in three true cases were missed, while specificity of 92.7% showed that most normal neonates were correctly identified as not having an ocular pathology by the red reflex. The negative predictive value of 85.7% confirmed that a normal reflex was highly reliable in ruling out disease. The positive predictive value of 76.6% highlighted the risk of false positives, which may cause parental anxiety but ensured that fewer true cases were missed. Existing literature affirm the findings of the current study, revealing strong specificity but variable sensitivity of the red reflex test Our results confirmed that the red reflex test was a valuable screening tool but not a definitive diagnostic test, and confirmatory ophthalmologic examination remained essential. Assisted vaginal delivery was strongly associated with ocular pathology. This finding was consistent with studies that linked mechanical birth trauma to retinal hemorrhage Most abnormalities resolved spontaneously, with 77.5% of neonates showing normal outcomes at follow-up. This can be explained by the fact that most pathologies identified, (retinal hemorrhages and transient immaturity in neonates) are known to spontaneously resolve over time Our findings support the integration of red reflex screening into routine newborn care in Kenya. This test is simple, inexpensive, and feasible for healthcare providers and it has great reliability in ruling out disease. Policy frameworks such as the Kenya National Eye Health Strategic Plan (2020-2025) emphasized early detection of childhood blindness but had limited implementation (Kenya Ministry of Health, 2020; PolicyVault The limited detection of rare but severe conditions such as congenital cataract and retinoblastoma highlight the need for innovative approaches
Conclusion
The red reflex test is a valuable and practical screening tool in a high-risk neonatal population. It demonstrated high accuracy and specificity and moderate sensitivity. Its high negative predictive value of 85.7% made it an excellent rule-out test, while its moderate positive predictive value necessitated confirmatory ophthalmologic evaluation. We advocate for the systematic adoption of red reflex screening in Kenyan neonatal units as a first step toward preventing avoidable childhood blindness through early detection of ocular pathologies and referral.