Open Globe Injury Repairs in the American Academy of Ophthalmology IRIS^® Registry 2014–2018: Incidence, Risk Factors, and Visual Outcomes

Abstract

Purpose:

To estimate incidence and evaluate demographic risk factors and visual acuity (VA) outcomes of open globe injuries requiring surgical repair in the IRIS^® Registry (Intelligent Research in Sight).

Design:

Retrospective cohort study.

Subjects, Participants, and/or Controls:

Patients with open globe injury repairs (OGR) were identified by Current Procedural Terminology codes (65275, 65280, 65285, 65286, 65235, 65260, 65265) from 2014 to 2018 in the IRIS Registry.

Methods Intervention or Testing:

Logistic regression models adjusting for age, sex, race, ethnicity, US region, concurrent and subsequent surgeries, and baseline VA.

Main Outcome Measures:

Outcomes included annual and 5-year incidence rates per 100,000 persons and factors associated with OGR, VA < 20/40, and VA ≤ 20/200 at final follow up (3 to 12 months post OGR).

Results:

A total of 13,766 OGR’s were identified during the study period, and 5-year incidence was 28.0/100,000. Annual incidence ranged from 11.1 to 12.7/100,000. OGR was associated with ages 21–40 years (yrs) compared to younger than 21 yrs (OR, 1.6 [95% CI: 1.5–1.7]); males compared to females (OR, 2.8 [CI: 2.7–2.9]); Black vs. White race (OR, 1.3 [CI, 1.2–1.4]); Hispanic vs. Non-Hispanic ethnicity (OR, 1.7 [CI, 1.6–1.8]); living in the South (OR, 1.4 [CI, 1.3–1.5]) and West (OR, 1.3 [CI, 1.2–1.4]) (vs Midwest) regions and inversely associated with Asian vs. White race (OR, 0.6 [CI, 0.6–0.7]). VA outcomes were analyzed in a subset of 2,966 patients with VA observations before the injury and at follow-up. Vision impairment (VA < 20/40) at final follow up was associated with VA ≤ 20/200 at presentation (20/200 better than 20/40 - OR, 11.1 [CI, 8.0–15.7]); older age e.g. 80+ yrs (vs < 21 yrs) (OR, 5.8 [CI, 3.2–10.7]); and Black race (vs White) (OR, 1.8 [CI, 1.3–2.6]). Risk factors were similar for VA ≤ 20/200 post OGR. Among the 1,063 OGR patients presenting with VA 20/200 or worse, VA did not improve to better than 20/200 at follow up in 35% (1,063/2,996) of patients.

Conclusions:

Our findings bring to light racial disparities in risk of OGR and poor visual outcomes that warrant further exploration.

Introduction

Ocular trauma is a significant cause of monocular visual impairment and blindness worldwide¹. Open globe injuries (OGI) are a severe type of ocular trauma in which there is a full-thickness laceration or rupture of the eye wall. These injuries often require surgery, can result in permanent visual impairment, and impose a significant economic burden on patients as well as on the health care system². Given that the majority of eye injuries are preventable, identifying populations at increased risk is essential in the development of strategies to aid in OGI prevention³.

Previous studies on OGIs have described the demographics⁴, causes of injury^{5, 6},epidemiology⁷, type and location of injury⁸,incidence⁹, visual outcomes¹⁰,associated health care costs², and disparities in discharge patterns¹¹. OGI occur more frequently in younger males and the elderly over 80 years of age^{4, 6, 8}, and in domestic accidents caused by blunt or sharp objects^{2, 5–7}. Poor visual acuity at presentation⁹, zone of involvement¹², and time to presentation¹² have been found to be predictive of visual outcomes.

Most of the previous literature has been limited to single center studies, and our understanding of OGIs at a national level is lacking. The United States Eye Injury Registry, an important resource for collecting self-reported multicenter data⁵, was created to circumvent this limitation but has not been active since 2013¹³. Our goal with this retrospective observational study was specifically to evaluate incidence, risk factors, and visual acuity (VA) outcomes of OGI requiring open globe repairs (OGR) throughout the United States. We leveraged data from the American Academy of Ophthalmology (Academy) IRIS^® Registry (Intelligent Research in Sight), a national database containing selected de-identified electronic health record data from more than 75 million unique patients as of July 2022¹⁴ to study OGR epidemiology and outcomes to estimate OGR incidence and evaluate demographic factors associated with the risk of OGR and the risk of visual impairment or legal blindness within 1 year following OGR.

METHODS

This study was conducted in accordance with the Declaration of Helsinki¹⁵. Given the use of deidentified patient data, this study was exempted from review by the Wills Eye Hospital Institutional Review Board.

Primary Data Source

The IRIS Registry was the source of data for this retrospective cohort study. The IRIS Registry contains de-identified electronic health record data (EHR) and is one of the largest clinical specialty data registries in the United States^{16, 17}. As of July 2022, the registry contained data on over 75 million patients and over 454 million patient visits from approximately 15,799 clinicians in ophthalmology practices in the US¹⁴. Version Chicago_amc_2021_04_16 last modified on April 16, 2021 was utilized for this analysis. Procedures for data collection and aggregation used for the IRIS Registry have been previously described¹⁸. Briefly, individuals are specified by unique randomly-generated identifiers and linked to demographic information, diagnoses, procedures, visual acuity (VA) and therapies. Data were queried from the IRIS Registry using PostgreSQL 8.0.2 (Portland, OR, USA), and all analyses were performed within the Amazon Web Services Virtual Private Cluster environment.

Study Population

All patients in the IRIS Registry who underwent OGR, identified by the Current Procedural Terminology (CPT) codes of 65275, 65280, 65285, 65286, 65235, 65260, and 65265 from January 1, 2014 to December 31, 2018 were considered for inclusion in these analyses. Patients missing data on age or gender were excluded.

Visual acuity outcomes

For inclusion in the analysis of visual acuity outcomes, patients were required to have distance VA measurements at presentation and during at least one follow-up between 3 months and one year after surgery. Patients missing VA data at presentation, or follow-up between 3 months and 1 year were excluded from the VA analysis cohort. Additional data processing steps were necessary to obtain the final VA analysis cohort. When procedure laterality was not available in the IRIS Registry, VA measurements on the day following OGR were used to impute procedure laterality. When VA was measured in a single eye the day after OGR, the procedure was inferred to be in that eye; otherwise the patient was dropped from the study. Patients with bilateral OGR were excluded from the study if they had left and right OGR procedures performed on separate dates. When the procedure was bilateral on a single date, one eye was randomly chosen using a random number generator for inclusion in the study. If an eye had multiple VA measurements on a single date, then one measurement was selected for inclusion based on VA metadata in the following order of preference: (1) best-corrected VA (refraction), (2) pinhole, (3) corrected, and (4) uncorrected. If there were multiple same-day VA measurements of the same type, the best VA was chosen to describe the best-case scenario for recovery from OGI. VA data were converted to LogMAR units from stored Snellen values by calculating the negative log of the Snellen fraction.

Study Outcomes

The outcome variables for this study were annual (average) and cumulative 5-year rates of OGR, visual impairment (VA < 20/40) or legal blindness (VA ≤ 20/200) within 3 to 12 months following surgery. Annual and 5-year incidence rates were calculated for each year as the number of OGR patients for that year divided by the number of patients in the IRIS Registry during the same period per 100,000. A patient was considered present for a specific year if that patient had at least one visit during that year. If a patient, for example, had 3 visits across 5 different years, then that patient would be counted only once for the cumulative incidence rate but the patient would also be counted once for each of the three years for which the annual incidence rates are calculated. Following this logic, we expect the cumulative incidence rate not be the additive sum of annual incidence rates because the number of patients over the 5-year period is not equal to the sum of the patients at each year.

The distribution of VA outcomes was described in terms of mean LogMAR at 5 distinct time periods: presentation (the day before or the day of OGR), 3 months (78 to 106 days after, closest to 92), 6 months (150 to 210 days after, closest to 183), 1 year (273 to 457 days after, closest to 365), and final follow-up (latest visit 92 to 365 days post OGR, approximately 3 months to 1 year). VA outcomes were also described with frequencies and percentages falling into discrete categories such as VA < 20/40 or VA ≤ 20/200.

Covariates

Sociodemographic (age [categorical], gender, race, ethnicity, and geographic location of residence) and procedure (surgery type, VA, concurrent and subsequent surgical procedures post OGR) data were collected. Concurrent and subsequent procedures from January 1, 2014 to December 31, 2018 were grouped in categories defined by the following CPT code sets: retinal [67036, 67039, 67040, 67041, 67042, 67043, 67108, 67113, 67107]; corneal [65755, 65730, 65750, 65756]; cataract removal [66850, 66920, 66982, 66984, 66840, 66930, 66940, 66983]; glaucoma-related [66150, 66155, 66160, 66170, 66172, 66183]; tube shunt [66179]; iris/ciliary body repair [66680]; anterior chamber washout [65800]; anterior vitrectomy [67005,67010]; iridectomy [66761]; eyelid laceration [67930,67935]; secondary intraocular lens (IOL) [66985, 66986, 66825]; silicone oil removal [67121, 65920].

Statistical Analyses

Demographic and clinical covariates were included in univariable and multivariable logistic regression models to estimate odds ratios (OR) and 95% confidence intervals. To evaluate factors associated with OGR, demographic covariates included age (years), gender, race, ethnicity, and a patient’s US geographic region of residence. To evaluate VA outcomes (VA < 20/40 and VA ≤ 20/200 at final follow-up) covariates included VA at presentation, presence of concurrent surgeries at the time of the OGR, and surgical procedures up to 12 months following OGR in addition to the demographic covariates. A two-sided P-value of <0.01 was considered statistically significant. Chi-square tests were used to compare the distribution of categorical variables between the VA analysis cohort and the cohort of patients excluded from the VA analysis. Statistical analysis and descriptive statistics were performed using R Version 3.6.0 (https://www.r-project.org/) and/or python (Python Software Foundation, http://www.python.org).

Results

OGR Incidence

A total of 13,766 patients with OGR between 2014 and 2018 with known gender and age were identified from the IRIS Registry. Annual incidence was similar across the 5 years ranging from 11.7 per 100,000 in 2014 to 11.1, 12.2, 12.7 and 12.3 per 100,000 between 2015 and 2018 respectively, with an average incidence of 12.0 per 100,000 per year. The 5-year cumulative incidence of OGR was 28.07 per 100,000 patients and varied with age, gender, race, ethnicity, and geographic region. Incidence was highest (44.6 per 100,000) in adults 21–40 years old and lowest (9.0 per 100,000) in adults over 80 years old (Table 1). Incidence was 3 times higher in males (44.5 per 100,000) than in females (15.7 per 100,000). Rates were higher in patients whose race was Black (39.5 per 100,000) or classified as Other (72.7 per 100,000) and lower in White and Asian patients (27.8 and 19.8 per 100,000 respectively). OGR incidence rates in patients with Hispanic ethnicity were almost double to the rates in patients who are Not Hispanic (49.3 and 28.2 per 100,000 respectively) (Table 1). A unimodal age distribution with a peak age for OGR of 21–40 years was seen in males, while incidence rates for females were relatively constant across age-groups (Supplemental Figure 1). Incidence also varied by geographic region, was highest in the South region (33.5 per 100,000), decreased to 30.3 per 100,000 in the West and lowest for Midwest and Northeast regions (22.4 and 23.3 per 100,000 respectively).

Table 1 –

Demographics of IRIS Registry Patients with and without Open Globe Repairs (OGR) and OGR Incidence (per 100,000) (2015–2019)

	OGR		IRIS (No OGR)		5-yr cumulative incidence/100,000
	%	N	%	N
Total	100	13,766	100	49,019,574	28.08
Age
0–20	8.80	1,212	8.19	4,015,834	30.18
21–40	21.74	2,993	13.68	6,705,643	44.63
41–60	28.51	3,924	21.39	10,487,702	37.42
61–80	36.66	5,046	42.35	21,251,839	23.74
>80	4.29	591	13.38	6,558,556	9.01
Gender
Female	31.93	4,396	57.12	28,001,567	15.70
Male	68.07	9,370	42.88	21,018,007	44.58
Race
White	63.38	8,725	64.01	31,375,335	27.81
Asian	2.18	300	3.08	1,508,725	19.88
Black or African American	10.58	1,456	7.51	3,683,512	39.53
Other	2.38	328	0.92	450,288	72.84
Unknown	21.48	2,957	24.48	12,001,714	24.64
Ethnicity
Not Hispanic or Latino	67.04	9,229	66.52	32,610,014	28.30
Hispanic or Latino	14.83	2,042	8.44	4,134,968	49.38
Unknown	18.12	2,495	25.04	12,274,592	20.33
Region *
Midwest	16.31	2,245	20.36	9,980,522	22.49
Northeast	17.41	2,397	20.91	10,247,982	23.39
South	44.54	6,132	37.33	18,300,931	33.51
West	19.69	2,710	18.22	8,929,836	30.35
Unknown	2.05	282	3.18	1,560,303	18.07

Abbreviations: OGR = open globe repairs; IRIS = Intelligent Research in Sight

^*Geographic Region where patients reside were as follows: Midwest (IN, IL, MI, OH, WI, IA, KS, MN, MO, NE, ND, SD); Northeast (CT, ME, MA, NH, RI, VT, NJ, NY, PA), South (DE, DC, FL, GA, MD, NC, SC, VA, WV, AL, KY, MS, TN, AR, LA, OK, TX); West (AK, CA, HI, OR, WA, AZ, CO, ID, NM, MT, UT, NV, WY).

OGR Associated Factors

Age, sex, race, ethnicity, and region were statistically significantly associated with OGR in both univariable and multivariable regression models when compared to patients without OGR (Table 2). Compared to patients <21 years old, the odds (OR (95% CI)) of undergoing OGR peaked in patients 21–40 years old (OR: 1.6, 1.5, 1.7)) and progressively decreased with increasing age with the lowest odds of 0.32 (0.29, 0.35) in patients over 80 years old. Males were almost 3 times more likely than females (OR: 2.8, (2.7, 2.9)) to undergo OGR. Compared to White patients, Black (OR: 1.3, (1.2, 1.4)) and Other race patients (OR: 2.3, (2.0, 2.6)) had increased odds of OGR; Asians were less likely (OR: 0.6, (0.6, 0.7)). OGR was more likely in Hispanic patients (OR: 1.7, (1.6, 1.8)) compared with Non Hispanic patients. The odds of OGR also varied with a patient’s geographic region; compared to patients residing in the Midwest, patients from the South (OR: 1.4, (1.3, 1.5)) or West (OR: 1.3, (1.2, 1.4)) regions had increased odds.

Table 2 –

Demographic Risk Factors for having an Open Globe Repair (N=13,766).

	Multivariate OR*	95% CI	p-value
Age
0–20 (Reference)	1	[Reference]
21–40	1.63	(1.53, 1.74)	<.001
41–60	1.36	(1.27, 1.45)	<.001
61–80	0.85	(0.8, 0.91)	<.001
81–98	0.32	(0.29, 0.35)	<.001
Birth Sex
Female (Reference)	1	[Reference]
Male	2.87	(2.77, 2.97)	<.001
Race
White (Reference)	1
Asian	0.67	(0.6, 0.76)	<.001
Black	1.37	(1.29, 1.45)	<.001
Other	2.34	(2.09, 2.61)	<.001
Unknown	0.79	(0.75, 0.83)	<.001
Ethnicity
Not Hispanic (Reference)	1	[Reference]
Hispanic	1.71	(1.62, 1.8)	<.001
Unknown	0.77	(0.73, 0.81)	<.001
Region
Midwest (Reference)	1	[Reference]
Northeast	1.03	(0.98, 1.10)	0.26
South	1.45	(1.38, 1.52)	<.001
West	1.36	(1.29, 1.44)	<.001
Unknown	0.72	(0.64, 0.82)	<.001

Abbreviations: OR= odds ratio; CI: confidence interval

^*Logistic regression model using age, gender, race, ethnicity, and region as covariates.

Visual Acuity (VA) Outcomes

Analyses on visual acuity outcomes within 1 year of OGR are based on the subset of patients (N=2,966 (21.5%)) who had VA observations the day of or the day before OGR and at least 1 VA observation between 3 months to 1 years post-OGR (Figure 1). Overall, VA improved by about three lines of vision within 1 year, from a mean LogMAR value of 1.26 at presentation to 0.89 at final follow up (Table 3). The proportion of eyes with visual impairment decreased by 25%, from 71% to 53%. Similarly, legal blindness decreased by one third, from 53% to 35%. Analysis of patients VA stratified by VA Snellen category at presentation suggests that VA stabilizes within the first three months post-OGR regardless of the VA at presentation (Supplemental Figure 3).

Figure 1.

Inclusion criteria for Open Globe Injury Repair (OGR) Visual Acuity (VA) Cohort.

Table 3.

Baseline and Follow-up Visual Acuity (VA) of Patients Undergoing Open Globe Injury Repair ^* (N = 2966 patients, 2966 eyes)

	Presentation 1		Month 3 2		Month 62		Year 12		M3Y12
	%	N	%	N	%	N	%	N	%	N
Patients	100	4,278	45.65	1,953	44.97	1,924	48.53	2,076	69.33	2,966
Eyes	100	4,496	45.20	2,032	44.73	2,011	48.62	2,186	65.97%	2,966
Visual Acuity (VA)
VA 20/40 or Better	28.54	1,283	34.69	705	37.54	755	43.09	942	46.97	1,393
[20/40, 20/200)	18.22	819	20.77	422	20.84	419	18.25	399	17.19	510
[20/200, CF)	12.12	545	13.63	277	14.32	288	12.03	263	10.59	314
CF	10.52	473	12.11	246	9.45	190	8.83	193	8.19	243
HM	16.93	761	11.42	232	9.15	184	9.33	204	8.36	248
LP	11.61	522	5.61	114	6.46	130	5.72	125	5.66	168
NLP	2.07	93	1.77	36	2.24	45	2.74	60	3.03	90
Worse than 20/40	71.46	3,213	65.31	1,327	62.46	1,256	56.91	1,244	53.03	1,573
20/200 or Worse	53.25	2,394	44.54	905	41.62	837	38.66	845	35.84	1,063

Abbreviations: SD= standard deviation; CF= counting fingers; HM= hand motion; LP=light perception; NLP= no-light perception

^*Based on all eyes with VA measurements at presentation.

¹Presentation VA is defined as measurements taken the day before or the day of OGR

²Month 3, 6, Year 1, and M3Y1 VA’s indicate times VA measurements were taken after OGR between 78–106, 150–210, 273–457, and 92–365 days respectively

Risk Factors for Visual Impairment (VA < 20/40) and Legal Blindness (VA ≤ 20/200) following OGR

Age, gender, race, ethnicity, VA at presentation, and number of subsequent surgeries were all associated with visual impairment within 1 year of OGR (Table 4). Compared to patients younger than 20 years old, the odds of visual impairment increased with increasing age, peaking in patients over 80 years old (OR: 5.8 (3.2, 10.7)). In contrast to OGR incidence, males were less likely to have visual impairment during follow up (OR: 0.7 (0.6, 0.9)) compared to females. Visual impairment was associated with being Black (OR: 1.8, (1.3, 2.6)) compared to being White, and with being Hispanic (OR: 1.4, (1.0, 2.0)) compared to being Non-Hispanic. Finally, poor VA at presentation was the strongest predictor of visual impairment; patients with 20/200 or worse at presentation were 10 to 30 times more likely to be visually impaired within 1-year post-OGR. Results for risk factors for legal blindness were similar to those for vision impairment (Table 5).

Table 4 –

Risk Factors for Visual Impairment (VA Worse than 20/40) in the VA Analysis Cohort^* (N = 2966)

	OGR		Multivariable OR	95% CI	p-value
	%	N
Age (years)
0 – 20 (Reference)	7.11	211	[Ref]
20 – 40	19.52	579	1.38	(0.94, 2.04)	0.10
40 – 60	28.19	836	1.89	(1.31, 2.77)	<0.001
60 – 80	40.22	1,193	2.67	(1.85, 3.86)	<0.001
>80	4.96	147	5.84	(3.26, 10.72)	<0.001
Gender
Female (Reference)	33.31	988	[Ref]
Male	66.69	1,978	0.78	(0.63, 0.95)	0.01
Race
White (Reference)	69.62	2,065	[Ref]
Asian	2.63	78	0.84	(0.48, 1.49)	0.55
Black	10.01	297	1.87	(1.35, 2.6)	<0.001
Other	2.43	72	1.41	(0.78, 2.59)	0.26
Unknown	15.31	454	1.01	(0.75, 1.36)	0.96
Ethnicity
Not Hispanic (Reference)	73.77	2,188	[Ref]
Hispanic	12.54	372	1.46	(1.07, 2.01)	0.01
Unknown	13.69	406	0.83	(0.62, 1.11)	0.20
Region
Midwest (Reference)	20.77	616	[Ref]
South	42.45	1,259	1.15	(0.9, 1.46)	0.27
West	21.98	652	1.35	(1.02, 1.79)	0.03
Northeast	13.12	389	1.38	(1, 1.91)	0.05
Unknown	1.69	50	2.02	(1, 4.17)	0.05
Visual Acuity at Presentation
Better than 20/40 (Reference)	16.76	497	[Ref]
[20/40, 20/200)	24.31	721	3.54	(2.65, 4.76)	<0.001
[20/200, CF)	13.18	391	11.18	(8.04, 15.7)	<0.001
CF	12.10	359	14.86	(10.49, 21.3)	<0.001
HM or Worse	33.65	998	34.29	(25.07, 47.46)	<0.001
Concurrent Surgery
No (Reference)	79.06	2,345	[Ref]
Yes	20.94	621	0.79	(0.64, 0.99)	0.04
Subsequent Surgery
No (Reference)	65.64	1,947	[Ref]
Yes	34.36	1,019	1.99	(1.62, 2.45)	<0.001

Abbreviations: VA= visual acuity; OGR= open globe repair; OR= odds ratio; CI= confidence interval; CF=counting fingers; HM=hand motions

^*The analysis cohort (N = 2966) is the group at the end of the data flow chart. Eyes included in the VA analysis cohort have age, birth sex, VA at presentation, and VA in the 3M1Y follow up period. One eye is randomly chosen from patients with bilateral procedures on the same day. All other eyes are excluded.

^**Odds ratios are high and confidence intervals are wide for the LP group in part because of small sample size

Table 5 –

Risk Factors for Legal Blindness (VA 20/200 or Worse) in the VA Analysis Cohort^* (N = 2966)

	OGR		Multivariable OR**	95% CI	p-value
	%	N
Age (years)
0 – 20 (Reference)	7.11	211	[Ref]
20 – 40	19.52	579	1.22	(0.8, 1.9)	0.35
40 – 60	28.19	836	1.61	(1.07, 2.44)	0.02
60 – 80	40.22	1,193	2.02	(1.36, 3.05)	<.001
>80	4.96	147	3.43	(2.01, 5.92)	<.001
Gender
Female (Reference)	33.31	988	[Ref]
Male	66.69	1,978	0.79	(0.66, 0.98)	0.02
Race
White (Reference)	69.62	2,065	[Ref]
Asian	2.63	78	1	(0.58, 1.73)	0.99
Black	10.01	297	1.61	(1.19, 2.16)	<.01
Other	2.43	72	1.17	(0.66, 2.09)	0.58
Unknown	15.31	454	0.88	(0.65, 1.19)	0.40
Ethnicity
Not Hispanic (Reference)	73.77	2,188	[Ref]
Hispanic	12.54	372	1.5	(1.1, 2.04)	0.01
Unknown	13.69	406	0.92	(0.68, 1.23)	0.55
Visual Acuity at Presentation
Better than 20/40 (Reference)	16.76	497	[Ref]
[20/40, 20/200)	24.31	721	2.42	(1.53, 3.96)	<.001
[20/200, CF)	13.18	391	9.34	(5.97, 15.18)	<.001
CF	12.10	359	15	(9.56, 24.42)	<.001
HM or Worse	33.65	998	33	(21.79, 52.35)	<.001
Concurrent Surgery
No (Reference)	79.06	2,345	[Ref]
Yes	20.94	621	0.74	(0.59, 0.92)	<.01
Subsequent Surgery
No (Reference)	65.64	1,947	[Ref]
Yes	34.46	1,019	1.45	(1.21, 1.75)	<.001

Abbreviations: VA= visual acuity; OGR= open globe repair; OR= odds ratio; CI= confidence interval; CF=counting fingers; HM=hand motions

^*The VA analysis cohort (N = 2966) corresponds to the last group in Figure 1. Eyes included in this cohort had available data on age, birth sex, VA at presentation, and VA in the 3M1Y follow up period. One eye was randomly chosen from patients with bilateral procedures on the same day.

^**The final logistic regression model was based on stepwise variable selection and included age, birth sex, race, ethnicity, visual acuity at presentation, concurrent and subsequent surgery as covariates.

Associated Factors: Concurrent and subsequent surgeries.

We examined the frequency and impact of additional (concurrent and/or subsequent) surgeries on VA outcomes following OGR. Previous studies reported that subsequent surgeries are associated with poor VA outcomes^{6, 19}, suggesting that more severe injuries might require supplementary surgeries and carry a worse prognosis. Here, frequency and type of concurrent and subsequent surgeries were analyzed (Supplemental Table 6). Overall, 26% of patients had additional surgeries simultaneously at the time of OGR and 25% of patients required subsequent surgeries (Supplemental Figure 4). In the VA cohort, patients undergoing surgeries subsequent to OGR were at higher risk of visual impairment (OR: 1.9, (1.6, 2.4)) compared to patients who did not require subsequent surgery (Table 4). Conversely, patients receiving multiple surgeries on the same day as the OGR were at decreased risk of visual impairment (OR: 0.7, (0.6, 0.9)) during follow up compared to patients not receiving additional surgeries.

Discussion

Open globe injuries are a significant cause of monocular blindness worldwide.

Using the IRIS Registry to examine incidence, visual acuity, and associated risk factors for OGR in the United States, this is the first study to document racial and ethnic disparities in both the risk of OGR and the risk of poor visual outcomes within 1-year post OGR. OGR incidence was highest in males, in patients 21–40 years old, and in Black and Hispanic patients. Our study results also were consistent with previous observations regarding poor visual acuity at presentation being predictive of final visual acuity post OGR^{4, 6, 10}.

Annual OGR incidence was stable across the five year study period, ranging from 11.1 to 12.7 per 100,000 in this national registry of ophthalmology patients, rates that are in line with a range of rates from 8 to 15 per 100,000 individuals estimated in previous studies from several countries including the US^{9, 20–24}. Males 21–40 years and individuals over 80 years old were the groups with highest and lowest incidence respectively (Supplemental Figure 1). While our study and at least one other studyobserved²², a unimodal age distribution with the peak age between 21–40 years, other studies reported a bimodal age distribution for open globe injuries with first peak in the younger patients 21–40 years old and the second peak in patients over 70 years old^{2, 9}. OGR occurred more frequently in males than females (3:1 ratio) in the IRIS Registry, consistent with estimates across studies using diverse data sources (i.e. national databases, hospitals, private clinics, and tertiary referral centers) and conducted in a range of geographic locations including Singapore⁹, Thailand²⁵, Turkey²⁶, India¹⁰, Mexico²⁷, Saudi Arabia²⁸, and the United States². Incidence rates also varied with race, ethnicity, and geographic region. The differences observed across studies may be due to differences in the definitions (i.e., open globe injury vs. repair), methods for sample selection (e.g., code-based vs clinical data), selection of the underlying population at risk (i.e., census data vs. IRIS Registry data), smaller sample sizes, or differences in geographic, racial, and ethnic variation in population studies. The national scope of patients in the IRIS Registry is more representative of patients seen in ophthalmic clinical practices in the US today and may offer insights with implications for public health relevance.

OGR incidence was associated with race, ethnicity, and geographic region in our study. Compared to White patients, OGR risk was higher in Black patients and patients of Other race and lower in Asian patients. Similarly, compared to Non-Hispanic ethnicity, OGR risk was higher in Hispanic patients. OGR risk was highest for those with residence in the South and West compared to the Midwest region; the reasons for this geographic variation remain unclear. Our observations of a higher OGR incidence in Black and Hispanic patients in the United States are consistent with some, but not all previous reports. A study by Tielsch et al using Maryland State Health Services Cost Review Commission data between 1979 and 1986, which classified race as White and Nonwhite, found that Nonwhites had higher OGI rates in the 25 to 69 age range²⁹. A more recent report using the National (Nationwide) Inpatient Sample (NIS) dataset between 2002 and 2013, found that OGI incidence was highest in young Black men and generally higher in Blacks, Hispanics and Native Americans compared to Whites in each age category⁴. These disparities may not have been captured by other studies because the underlying population was less racially or ethnically diverse, the sample size was small¹⁹, or data on race and ethnicity was not available². Although the cause of injury is not available in the IRIS Registry, the observation that Black and Hispanic patients are at higher risk for OGR when compared to White or Non-Hispanic patients, and that Asian patients are at lower risk is consistent with previous studies ^{30, 31}.

Vision impairment (VA < 20/40) and legal blindness (VA ≤ 20/200) at 1-year post OGR were strongly associated with increasing age (Tables 4, 5). Visual acuity at presentation varied with age; about 50% of patients between ages 20 to 40 years presented with a VA of HM or worse compared to over 60% of patients aged 60 or older (Supplementary Figure 5). The analysis on risk factors for OGR and risk factors for poor visual outcomes suggests that the relative odds of OGR decreasing with increasing age, while the odds of a poor VA outcome post OGR increases. This pattern is consistent with previous studies⁵ demonstrating in elderly patients a higher incidence of fall-related blunt ruptures which tend have poorer VA prognosis, compared to focal lacerations in younger patients such as from work-related projectile objects. In the IRIS registry, males were slightly less likely to have a poor VA prognosis, both for vision impairment and legal blindness.

Being Black or Hispanic was not only associated with higher risk for OGIs, but was also associated with poor VA post OGR. A series of studies on eye injuries in the Baltimore area by Katz, Tielsch, Sommer et al observed that although the injury rates among Black and White men were not different, the visual consequences of the injuries were more severe for Black men^32–34. Other retrospective cohort studies of 6,821 and 27,467 patients respectively with a primary diagnosis of OGI in the National Inpatient Sample (NIS) also found racial disparities but did not directly assess visual outcomes^{4, 7}. Our findings warrant further investigation using more granular clinical data.

Finally, in the IRIS Registry, having additional surgeries concurrently or subsequently to OGR were associated with poor VA within one year. Specifically, having concurrent surgeries decreased the risk of vision impairment or legal blindness post OGR by 20%, while having subsequent surgeries doubled the risk of vision impairment. Our findings about subsequent surgeries are consistent with previous studies^{6, 8, 19} and may be because more severe injuries require a greater number of subsequent surgeries. We found that having concurrent surgeries was not associated with visual impairment as an outcome and was inversely associated with legal blindness. One possible explanation is that eyes with less severe injuries have a better view at the time of open globe injury repair for concurrent procedures such as a traumatic cataract removal or retinal detachment repair. Given the limitations in ascertaining the timing of additional diagnoses such as traumatic cataract and retinal detachment in clinical registries, further studies regarding optimal timing of additional surgeries are warranted.

Our study had several limitations. Given the IRIS registry is a clinical registry, the accuracy of the data depends on coding accuracy and can be error prone. To improve the validity of the diagnosis, we identified patients who had OGR identified by CPT codes, rather than OGI using ICD-10 codes, selecting those cases requiring surgical repair. Our approach, while possibly underreporting some OGR patients, overall would capture significant OGIs. since most OGIs require surgical repair In our study, we were unable to define the causes of injury as a patient’s record in IRIS Registry has limited granularity and OGI zone or mechanism of injury are not available. Another important limitation lies in the underlying distribution of providers who contribute data to the IRIS Registry, and the mix of practitioners may change over time. Furthermore, the IRIS Registry currently includes only about one third of large academic centers and does not include Emergency Department data¹³, thus limiting the generalizability and availability of clinical data on presentation.

Despite these limitations, the IRIS Registry, with de-identified records on over 75 million patients, remains a rich and diverse source of data on demographics, diagnoses, and surgical procedures in a real-world setting. Given it is the largest clinical registry in the world, the IRIS Registry is ideal for studying less common conditions such as open globe injuries. The strength of the IRIS Registry is highlighted by its use to date as a data source for more than 50 publications, covering a wide variety of clinical, public health and health disparity topic as well as investigation of rare conditions and clinical outcomes.

Conclusions

Our results highlight racial and ethnic disparities in risk of OGR and, importantly, poor visual outcomes within 1-year post OGR. Additionally, our study confirms previously established risk factors associated with OGR for younger age and male sex and supports demographic-specific differences in high-risk injury settings. These findings combined with previous observations may offer guidance for targeting measures to prevent OGIs and associated vision loss.

Abbreviations:

OGI

Open globe injury

OGR

open globe injury repair

VA

visual acuity

IRIS

Intelligent Research in Sight