Causes of Childhood Blindness in the United States using the IRIS^® Registry (Intelligent Research in Sight)

Abstract

Purpose:

To investigate causes of childhood blindness in the United States using the IRIS^® Registry (Intelligent Research in Sight).

Design:

Cross-Sectional Study.

Participants:

Patients ≤18 years of age with visual acuity 20/200 or worse in their better seeing eye in the IRIS Registry during 2018.

Methods:

Causes of blindness were classified by anatomical site and specific diagnoses.

Main Outcome Measures:

Percentages of causes of blindness.

Results:

Of 81,164 children with 2018 visual acuity data in the IRIS Registry, 961 (1.18%) had visual acuity 20/200 or worse in their better-seeing eye. Leading causes of blindness were retinopathy of prematurity (ROP) in 301 (31.3%), nystagmus in 78 (8.1%), and cataract in 64 (6.7%) patients. The retina was the leading anatomic site (47.7%) followed by optic nerve (11.6%) and lens (10.0%). A total of 52.4% of patients had treatable causes of blindness.

Conclusions:

This analysis offers a unique cross-sectional view of childhood blindness in the US using a clinical data registry. More than one-half of blind patients had a treatable cause of blindness.

Précis

The leading cause of childhood blindness in the United States is retinopathy of prematurity, and half of cases of childhood blindness are treatable.

Introduction

Childhood blindness places a significant burden on both families and society because of the long duration of a blind child’s visual disability.^1,2 In addition, visually impaired children often have delayed motor, language, emotional, social and cognitive development.³ The causes of childhood blindness vary based on the Human Development Index (HDI) measures, with high HDI countries reporting cortical visual impairment (CVI) as the leading cause and low HDI countries reporting cataract, inherited retinal diseases, and congenital abnormalities as leading causes.⁴ In low HDI countries, childhood blindness has also been shown to reduce life expectancy.¹ Due to the significant socioeconomic impact of childhood blindness, the World Health Organization (WHO) global initiative and action plan has placed a high priority on understanding and seeking to eliminate childhood blindness.⁴

Identifying treatable causes of childhood blindness allows resources to be directed to early screening and treatment of these diseases. However, only a few studies have focused on identifying the causes of childhood blindness in the United States. Possible reasons for the paucity of studies on this topic include the absence of a national registry of the blind and visually impaired and the difficulty of obtaining reliable population-based data on this topic.^5,6 As a result, most studies of childhood blindness in the United States have surveyed schools for the blind. While this is a convenient way to assess childhood blindness, these studies are unlikely to accurately represent the population-level breadth and depth of blinding disease in children because only about 13% of legally blind children attend schools for the blind in the United States. Surveys targeting schools for the blind thus inherently under-represent unenrolled children, with those who do not live nearby being particularly at-risk.⁷ Moreover, these studies do not consistently report data using WHO definitions of blindness.

The IRIS^® Registry (Intelligent Research in Sight) is the nation’s first comprehensive eye disease clinical registry.⁸ It allows the causes of blindness to be identified for children examined by ophthalmologists because data elements include not only diagnoses but also visual acuity. We hypothesized that causes of childhood blindness in the United States identified from the IRIS Registry will align with the causes identified and reported by other high HDI countries.

Methods

This was a cross-sectional study performed using electronic medical record data of patients from the IRIS Registry, Rome 2.0 dataset extract. The IRIS Registry includes clinical data from ophthalmologists and ophthalmology-affiliated optometrists in participating United States ophthalmic practices, with records primarily derived via direct extraction from practice electronic health records. As of July 2022, the Registry contains data on over 75.4 million unique patients, cared for by 15,799 clinicians in ophthalmology practices. It is the largest specialty clinical data registry in the United States.⁹

A limited de-identified version of the IRIS Registry was made available to approved academic research centers (IRIS Registry Analytic Center Consortium). The initial version made available to the Consortium was codenamed “Rome;” for this analysis we used version 2.0, frozen in August 2021. Available data included demographics, code-based diagnosis and procedure information, visual acuity, intraocular pressure measurements, and provider practice location (zip code). The investigation was approved by the Stanford Institutional Review Board. The research adhered to the Declaration of Helsinki and the United States Health Insurance Portability and Accountability Act. Data evaluated from the IRIS Registry are de-identified via expert determination, and thus the study did not require patient or parentlevel consent.

The study included patients ≤18 years of age with visual acuity data present in the IRIS Registry between January 1, 2018, and December 31, 2018. We limited our analysis to patients who have visual acuity (VA) of logMAR ≥1.0 (approximately 20/200) in their better-seeing eye. All diagnoses associated with these patients in their electronic health records (encounter diagnoses, special testing obtained, patient counseling) were included.⁸

Baseline demographics were obtained, including age, sex, and race/ethnicity as documented in practice electronic health records. Race and ethnicity data was based on self-reported classifications provided by patients, documented in ophthalmic practices’ medical records, and reported in the IRIS Registry. Reported categories for race available in the IRIS Registry dataset respectively were Black, Asian, white, American Indian/Alaskan Native, Native Hawaiian/Pacific Islander, Other, and Unknown (unrecorded), and reported categories for ethnicity were Hispanic or Latino, non-Hispanic or Latino, and unknown (unrecorded). We classified geographic region based on US Census Regions by provider/practice office zip code (West, Northeast, South, and Midwest). Urban and rural population density was similarly determined based on provider office zip code classification in the Federal Office of Rural Health Policy accessed March 30, 2021.¹⁰

On the basis of the 2018 WHO classification of vision impairment,¹¹ we collected data on children considered to be legally blind. The International Classification of Diseases 11 (2018) defines blindness as an individual with VA of worse than 3/60 (equivalent to 20/400) in the better eye and severe vision impairment as VA worse than 6/60 (equivalent to 20/200) in the better eye. We utilized corrected visual acuity measurements that were converted to logMAR units for analysis. Snellen VA, counting fingers, hand movement, and light perception were assigned logMAR values, but other notations such as blink to light or fix and follow were not coded in the IRIS Registry and thus not included in the analysis. We utilized each patient’s best-recorded visual acuity with correction at their most recent clinical examination. We included children with a best-recorded distance VA in the better-seeing eye of logMAR ≥1.0. Children meeting our blindness criteria who only had unspecified VA, near VA, or uncorrected VA measurements were manually reviewed because in cases of severe visual impairment optical correction may not be helpful.

All diagnoses attached to a patient in the electronic health records were included, and from these we inferred the underlying diagnosis of the cause of blindness, based on International Classification of Diseases, Ninth and Tenth Revision Clinical Modification codes (Table S1). Causes of blindness were classified by the anatomical site and the underlying diagnosis. In children having two or more diagnosis codes associated with blindness, the primary cause of visual loss was established by expert opinion based on a disease hierarchy agreed upon by two ophthalmologists (SRL, HWL) in consultation with each other (Table 2). Cell sizes less than 10 were suppressed consistent with the IRIS Registry cell size suppression policy.

Table 2.

Hierarchy of diseases causing childhood blindness

Leber congenital amaurosis

Albinism

Stargardt disease

Retinal dystrophy

Retinitis pigmentosa

Retinopathy of prematurity

Cerebral visual impairment

Coloboma, Aniridia

Optic nerve atrophy, Optic nerve hypoplasia

Retinal detachment without retinopathy of prematurity

Cataract, Aphakia

Corneal dystrophy

Removed/disorganized

Phthisis

Microphthalmos

Retinoblastoma

Keratoconus

Corneal scar

Glaucoma

Uveitis, Chorioretinitis

Nystagmus

Amblyopia

Results

A total of 81,164 children were identified in the IRIS Registry who met study inclusion criteria with available visual acuity data during the study period from January 1, 2018, to December 31, 2018. Of these children, 961 (1.18%) had a corrected VA of logMAR ≥1.0 in their better-seeing eye. Characteristics of blind children are shown in Table 3. The median age was 9 years.

Table 3.

Characteristics of patients with childhood blindness in the IRIS Registry

	IRIS Registry Patients ≤ 18 years of age with Legal Blindness (visual acuity 20/200 or worse in better-seeing eye) N = 961	All IRIS Registry Patients ≤ 18 years of age N = 7,424,616	p-value
Age (years)	N (%)	N (%)	<0.0001
0–2	282 (29.3)	410 320 (5.5)
3–5	70 (7.3)	833 628 (11.2)
6–9	159 (16.5)	1 581 105 (21.3)
10–18	450 (46.8)	4 599 563 (62.0)
Mean	8.49	10
Median	9	11
Range	18	18
SD	6.35	4.85
Sex			<0.0001
Male	527 (54.8)	3 578 756 (48.2)
Female	430 (44.7)	3 803 755 (51.2)
Not Reported	4 (0.42)	42 105 (0.6)
Race			<0.0001
Asian	23 (2.39)	215487 (2.90)
White	571 (59.41)	3378052 (45.49)
Black or African American	109 (11.34)	582652 (7.84)
Native Hawaiian and Other Pacific Islander	-	16789 (0.23)
Native American and Alaska Native	-	39926 (0.54)
Other	10 (1.04)	197426 (2.66)
Unknown	248 (25.80)	2994284 (40.33)
Ethnicity			<0.001
Hispanic or Latino	290 (30.17)	881137 (11.86)
Not Hispanic or Latino	420 (43.70)	3422224 (46.09)
Unknown	251 (26.11)	3121255 (42.03)
Urban/Rural			< 0.001
Urban	623 (64.8)	2 112 167 (28.4)
Rural	49 (5.1)	4 147 607 (55.9)
Unknown	289 (30.1)	1 164 842 (15.7)
Geographic Region			<0.0001
Northeast	34 (3.5)	1 335 802 (18.0)
South	450 (46.8)	2 657 099 (35.8)
Midwest	98 (10.2)	1 327 869 (17.9)
West	44 (4.6)	1 049 714 (14.1)
Missing	335 (34.9)	1 054 132 (14.2)

Compared to the IRIS Registry pediatric patient population at large, blind children were more likely to be younger in age, particularly 0–2 years old, and male (54.8% male and 44.7% female, respectively, compared to 48.2% male and 51.2% female among all pediatric patients; p<0.0001). The proportion of children who were Hispanic and, to a lesser degree, Black, was notably higher among blind children than in the pediatric population at large (30.2% versus 11.9% Hispanic, 11.3% vs. 7.8% Black; p<0.0001). Blindness was more common among children seen in urban than rural practices participating in the IRIS Registry (64.8% urban practices among blind children, compared to 28.4% among all pediatric patients; p<0.0001). Although there was more missing geographic data for blind children, predominance of the South as the most frequent geographic region was strikingly more pronounced for blind children compared to the pediatric population at large (46.8% vs. 35.8%, p<0.0001).

Causes of blindness in children

The leading diagnoses associated with childhood blindness (Table 4) were retinopathy of prematurity (ROP) (31.3%), nystagmus (8.1%), cataract (6.7%), and optic nerve atrophy (6.5%). CVI accounted for 2.4% of cases.

Table 4.

Causes of childhood blindness in the IRIS registry grouped by anatomical site

Anatomical site or sites affected	Number	% of total
Whole globe	73	7.6
Glaucoma/ Buphthalmos†	56	5.8
Microphthalmos	10	1.0
Phthisis	<10*	<1*
Removed/disorganized	<10*	<1*
Cornea	43	4.5
Keratoconus	25	2.6
Corneal scar†	15	1.6
Corneal dystrophy/ Other	<10*	<1*
Lens	96	10
Cataract†	64	6.7
Aphakia†	32	3.3
Uvea	43	4.5
Aniridia	19	2.0
Coloboma	15	1.6
Uveitis†	<10*	<1*
Retina	458	47.7
ROP†	301	31.3
Retinal detachment w/o ROP†	47	4.9
Albinism	28	2.9
Stargardt disease	28	2.9
Retinitis pigmentosa	26	2.7
Leber congenital amaurosis	13	1.3
Retinal dystrophy	11	1.1
Chorioretinitis	<10*	<1*
Retinoblastoma	<10*	<1*
Optic nerve	111	11.6
Optic nerve atrophy	62	6.5
Optic nerve hypoplasia	49	5.1
Cerebral visual pathway	23	2.4
Cortical visual impairment	23	2.4
Others	114	11.9
Nystagmus	78	8.1
Amblyopia†	36	3.7

^†Causes that are Treatable.

^*Cell sizes less than 10 were suppressed in compliance with the IRIS Registry cell size suppression policy.

Classified according to the affected anatomical site, we found that a plurality of blindness cases were associated with retina pathology (47.7%), followed by optic nerve (11.6%) and lens (10.0%). Among children with disorders of the retina, almost two-thirds (65.7%) had ROP, and the remainder had a retinal detachment, albinism, Stargardt disease, retinitis pigmentosa, Leber congenital amaurosis, retinal dystrophy, or chorioretinitis. Only one child in our data sample was presumed blind as a result of retinoblastoma.

A total of 52.4% of patients had diagnoses that are treatable. The most common treatable causes of blindness were ROP, retinal detachment, cataract, glaucoma, corneal scar, and amblyopia.

Discussion

In this cross-sectional study of more than 80,000 children in 2018 using the IRIS Registry, we identified 961 children with legal blindness. The most common presumed causes of blindness identified were ROP, nystagmus, and cataract. Our results were striking for their lack of conformity with other high HDI countries, including greater representation of ROP and cataract. More than one-half of the children had treatable diagnoses and nearly as many had retina-associated blindness. ROP was the most common treatable cause of blindness observed in this cohort.

Our study differed from other reports of childhood blindness in the USA because it was based on data from the IRIS Registry as opposed to survey-based data. Kong and colleagues⁶ reported the leading causes of childhood blindness in the USA were CVI (18%), optic nerve hypoplasia (15%) and ROP (14%) based on surveying 16 schools for the blind. In contrast we report a much lower incidence of CVI (2.4%) and optic nerve hypoplasia (5.1%) and a higher incidence of ROP (31%). These differences may reflect differences in the type of children who attend schools for the blind versus undergo an office-based examination by an ophthalmologist or changes in the causes of childhood blindness in the USA over the past decade. A population-based survey (e.g. census-based) would be a more accurate way to assess childhood blindness, but such a resource is not currently available; it’s feasibility would be limited by cost and logistics and depending on survey methodology, may rely on self-reported blindness etiology and/or lack of formal visual acuity testing.

The IRIS Registry includes data from the majority of ophthalmology practices in the USA,^38,39 and provides broad geographic representation.¹² We found ROP to be the most frequently associated cause of blindness in our study. In total, 301 patients (31.3%) in our study were presumed blind from ROP. Although direct comparison with our results is difficult, many other countries have reported ROP as the most common cause of childhood blindness.^15,16 Surprisingly, the percentage of children blind from ROP in our study is much higher than that reported from the UK and previous studies from the USA. The view that ROP is becoming less of a problem in high-income countries may be overstated. This may reflect increased survival rate of very low birth weight infants that are at greater risk of developing severe ROP. There are multiple studies from the USA reporting increased rates of Stage 4 and 5 ROP.^17–21 Another confounding factor is the focus of clinical trials on the prevention of adverse anatomic outcomes—posterior retinal detachment, retinal fold involving the macula, or retrolental tissue. All major clinical trials have focused on anatomic outcomes as proxies for success, including recent anti-vascular endothelial growth factor (anti-VEGF) trials, and this focus may have biased the perception that treatment prevents blindness when in reality it prevents anatomic complications (e.g. retinal detachment).^22–27

There is marked variation in the leading causes of pediatric blindness within and among geographic regions.³ For children in low HDI countries, cataract, retinal dystrophies, and congenital anomalies have been reported to be the most common causes of childhood blindness.¹³ In high HDI countries, CVI and optic nerve anomalies have been reported to be the most common causes of blindness.^6,14 Similarly, a national population-based epidemiological study in the United Kingdom reported the leading causes of childhood blindness were CVI (60.7%), followed by disorders of the optic nerve, and retinal disorders.¹⁴ Only 4.1% of the children in United Kingdom study were reported to be blind secondary to ROP. One possible reason for the low proportion of CVI cases in our cohort might be that participating ophthalmologists diagnosed children as having nystagmus or amblyopia instead of CVI may not have had access to the complete medical record and as a result the ophthalmologist may have been unaware of other non-ocular diagnoses. It may also be that children with CVI have other disabilities that preclude them from having their visual acuity tested. Lastly, some children diagnosed with CVI have visual acuity better than logMAR 1.0 and would therefore not be captured in this analysis.^28–30

We observed regional geographic variation even within this analysis. Although practices in the Southern Census region represented a plurality for the IRIS Registry pediatric population at large as well as the subset with childhood blindness, we noted the Southern region to be predominant among children with legal blindness (46.8%, with 10% or fewer patients in each of the other US Census regions and the balance reflected by a higher proportion of missing geographic data). It is possible that this reflects differences in neonatal risk factors, regional differences in number of at-risk pre-term low birth weight infants, and/or differences in ROP prevention and treatment practices, however, these findings warrant further study.

Other notable findings in our analysis that bear further research are racial/ethnic distribution of blind children relative to the broader IRIS Registry pediatric population. We noted a higher proportion of Hispanic patients and (to a lesser extent) Black patients among blind children (22.4% vs. 4.5%, and 11.3% vs. 7.8%, respectively) compared to all IRIS Registry pediatric patients. Further research is also needed to investigate differences, taking into account factors such as geographic location and associated diagnoses. Certain types of blindness may be overrepresented in different ethnic groups, also affected by factors such as access to health care.

Overall, more than half of the children enrolled in our study had treatable causes of blindness including ROP, cataract, glaucoma, and retinal detachment. This is in contrast to studies reporting CVI and disorders of the optic nerve as the most common causes of childhood blindness (conditions which are generally not treatable).^6,13 Although some of these patients may have received treatment and still have had poor visual outcomes, there is a potential opportunity for intervention since many of these conditions have available treatments. Public health measures should be taken to reduce the burden of blindness from these disorders, including measures such as education of healthcare professionals and family members, implementation of screening programs that adhere to the Joint Statement Screening Guidelines for ROP,³¹ and/or development and deployment of scalable artificial intelligence-based telehealth screening. In one recent survey of neonatal intensive care units (NICUs) on ROP practices, 20% of the NICUs were not using weight-based screening criteria despite three consecutive American Academy of Pediatrics and American Academy of Ophthalmology Screening Statements for ROP recommending this practice.³² This same study indicated that there was widespread belief (e.g. 99%) that no eligible infants were missed, yet large scale review of practice patterns for ROP screening within the Vermont Oxford Network discovered that between 7.5–16% of infants are not being screened in a timely manner.³³

Although ROP is classified as a treatable cause, present treatments focus on anatomical improvement; despite these improvements, the visual prognosis can be poor. While more effective treatments for ROP have been developed in recent years such as laser photocoagulation and intravitreal injection of anti-VEGF agents, ROP continues to be an important cause of childhood blindness in the USA.¹⁸ To prevent blindness from ROP, more attention to screening and treatment is warranted, but this problem is aggravated by the proliferation of NICUs (>1350 in the USA)³⁴ and the declining population of screeners.³² Early diagnosis and treatment is also essential for pediatric cataracts, the second most common treatable cause of blindness, because a delay in treatment can result in severe deprivation amblyopia.^35,36 Some of these delays arise from referrals from primary care doctors to pediatric ophthalmologists who do not perform cataract surgery.³⁶ Visual outcomes are closely related to the timing of cataract surgery.³⁷

There are several limitations to our study. First, we lack secondary information on causes of nystagmus and amblyopia from lack of access to the complete medical record. It is possible and even probable that many patients from the pre-term population have reduced vision from both ROP and CVI and thus it is difficult to determine the primary cause of blindness. Second, in some cases this data represents an evaluation of a child’s vision in a single point in time and is reliant upon recorded diagnosis codes. Third, there is no longitudinal follow-up to evaluate efficacy of interventions or improvement/worsening in care. Fourth, we do not have the ability to tease out whether blindness and/or severe visual impairment are disease-related or refraction related. Finally, blindness due to trauma or injury could not be determined because the diagnosis code is based on ICD-9/10 codes for which identifying trauma or injury is imperfect.

Despite these limitations, this paper highlights that approximately 50% of pediatric blindness in the USA is likely treatable and is associated with retinal pathology. This should refocus our attention to what we can identify, treat, and prevent. We know that early intervention can reduce adverse outcomes and that continuing functional success requires access to eye care providers for refractive and amblyopic management.