Abstract
Purpose
To evaluate the long-term predictive potential of the Retinal Detachment after Open Globe Injury (RD-OGI) Score, designed to predict the risk of RD after penetrating ocular trauma.
Patients and Methods
Retrospective cohort study at a single tertiary-care academic hospital with a level 1 eye trauma center (Wilmer Eye Institute, Johns Hopkins University) including patients presenting with acute open globe injury (OGI) between 2006 and 2017. Visual acuity (VA), zone of injury (ZOI), and presence of vitreous hemorrhage (VH) at presentation were utilized to calculate the RD-OGI score. Incidence of RD following OGI up to 720 days was compared among low-, moderate-, and high-risk groups stratified by RD-OGI score.
Results
A total of 597 eyes with OGI were stratified into low (n=165, 27.6%), moderate (n=254, 42.5%) and high risk (n=178, 29.8%) groups per the RD-OGI scoring system. Individuals in the high-risk group were significantly older compared with those in moderate and low risk groups (47.4±24.2, 40.2±21.4 and 34.4±17.7, respectively) (p<0.001). The most common highest ZOI for low and moderate risk groups was zone I [64.2% (106 out of 165 total) of low-risk eyes and 51.2% (130 out of 254 total) of moderate risk eyes)]. Most eyes in the high risk group had a highest ZOI of either zone II (52.8%, 94 out of 178 total) or zone III (36.5%, 65 out of 178 total). Of the 306 eyes with follow-up information available 720 days after OGI, 150 (49.0%) had developed RD, with majority of these patients (83 out of 150, 55.3%) belonging to the high risk group. The RD-OGI score exhibited its strongest predictive performance at 6 months, with an area under the curve (AUC) of 0.774 (95% confidence interval [CI]: 0.730–0.819), and demonstrated a slight decline in predictive performance by 720 days post-injury (AUC: 0.739, 95% CI: 0.684–0.794).
Conclusion
RD-OGI score provides a reliable method to stratify risk of developing RD after penetrating ocular trauma, with moderate predictive performance at 6 months (AUC: 0.774, 95% CI: 0.730–0.819), that was maintained over longer term follow up at 2 years (AUC: 0.739, 95% CI: 0.684–0.794).
Keywords: ocular trauma, open globe injury, retinal detachment, risk stratification
Introduction
Open globe injuries (OGI) represent a relatively uncommon yet significant cause of visual loss, with an estimated incidence of 4.49 injuries per 100,000 population in the US alone.1 Despite anatomical correction through surgical management, visual outcomes following OGI remain guarded.2 Several factors may influence visual prognosis, including complications such as cataract, secondary glaucoma, and retinal detachment (RD).3 The prevalence of RD at time of presentation ranges from 7.8% to 25.5% in eyes with OGI and is often associated with poor final vision.4–6 Importantly, even if the retina is attached at presentation, OGI can confer an increased risk of subsequent RD development that is highest early on, with a recent analysis reporting RD development within 90 days of OGI in 41.7% of eyes.7,8 Given the significant impact of RD and timing of RD repair on visual outcomes of these patients, identifying patients at high risk is critical to guide management.9
Recently, a scoring system (RD-OGI score) has been developed to predict the risk of RD following OGI.10 This scoring system assigns points to three presenting clinical variables: visual acuity (VA) [VA better than CF = 0 points, CF at 2 feet = 1 point, hand motions (HM) at 2 feet = 2 points, light perception (LP) = 2.5 points, no light perception (NLP) = 3.5 points], zone of injury (ZOI) defined by highest ZOI for a given eye [Zone I injury = 0 points, Zone II injury = 0.5 points, Zone III injury = 2 points], and presence of vitreous hemorrhage (VH) [VH = 2 points].10 The sum of points accrued for an eye with OGI (range 0 to 7.5) is then associated with an estimated probability of developing RD following OGI (range 1% for RD-OGI score of 0 to 95% for RD-OGI score of 7.5).10 This system holds potential to aid clinicians in stratifying risk and optimizing care. However, its effectiveness has yet to be evaluated on large external datasets with follow up longer than one year.11,12 Although the overall incidence of RD declines with time following OGI, validation of the RD-OGI scoring system with long-term follow-up would provide additional insight into risk of late RDs which may be more likely associated with proliferative vitreoretinopathy (PVR) and worse visual outcomes.13,14 The current study aims to validate the RD-OGI score as a prediction tool using an external cohort of patients from a single tertiary university hospital with a high volume of OGI cases with up to two years of follow up, assessing its generalizability and long-term clinical utility.
Materials and Methods
Data Collection
A retrospective chart review was conducted on patients treated for OGI at the Wilmer Eye Institute, Johns Hopkins University, between March 2006 and April 2017. The study was conducted in accordance with the tenets of the Declaration of Helsinki and approved by the Johns Hopkins Medicine Institutional Review Board (IRB00132759). Due to the retrospective nature of the study and minimal risk involved to the privacy of individual participants, a waiver for written informed consent was approved by the Johns Hopkins Medicine IRB. All data were deidentified and kept confidential. The data extracted from electronic medical records included visual acuity (VA) at time of presentation, zone of injury (ZOI), presence or absence of vitreous hemorrhage (VH), time to RD, and time to the last follow-up. VA was recorded as Snellen and converted to logMAR equivalents. The ZOI was classified according to the Ocular Trauma Classification Group criteria as zone I (injury confined to the cornea), zone II (injury involving the sclera up to 5 mm posterior to the limbus), or zone III (scleral injury more than 5 mm posterior to the limbus).15 Exclusion criteria were lack of the necessary information in the chart to calculate RD-OGI score (VA, ZOI, presence/absence of VH) and less than 30 days between initial presentation and final follow-up visit. Individuals in the latter category were considered as lost to follow up and excluded to minimize confounding of long-term outcomes data. Time to RD was defined as the interval from initial surgical repair to the confirmed diagnosis of RD by the attending physician. RD was confirmed through fundus examination when visible or by B-scan ultrasonography or CT imaging when the fundus could not be visualized. RD-OGI score was calculated for each patient (Table 1), and patients were then stratified into three risk groups according to their total score – low risk (0–2 points), moderate risk (2.5–4.5 points), and high risk (5–7.5 points).10,11 This stratification was used for the final analysis to assess the validity of the RD-OGI score in predicting RD risk.
Table 1.
Distribution of Age, Ethnicity, Sex and Affected Eye Among All Individuals in Low Risk, Moderate Risk and High Risk RD-OGI Groups
| All (n=597) | Low-Risk (n=165) |
Moderate-Risk (n=254) |
High-Risk (n=178) |
p-value | ||
|---|---|---|---|---|---|---|
| Age at presentation (in years, mean±SD) | 40.7 ± 21.9 | 34.4 ± 17.7 | 40.2 ± 21.4 | 47.4 ± 24.2 | <0.001* | |
| Ethnicity/Race† (n, %) | Caucasian | 325 (54.6%) | 91 (55.8%) | 134 (52.8%) | 100 (56.2%) | 0.24 |
| African-American | 163 (27.4%) | 40 (24.5%) | 67 (26.4%) | 56 (31.5%) | ||
| Hispanic | 87 (14.6%) | 28 (17.2%) | 41 (16.1%) | 18 (10.1%) | ||
| Other | 20 (3.4%) | 4 (2.5%) | 12 (4.7%) | 4 (2.2%) | ||
| Sex (n, %) | Female | 131 (21.9%) | 33 (20.0%) | 51 (20.1%) | 47 (26.4%) | 0.23 |
| Male | 466 (78.1%) | 132 (80.0%) | 203 (79.9%) | 131 (73.6%) | ||
| Eye affected (n, %) | OD | 273 (45.7%) | 68 (41.2%) | 127 (50.0%) | 78 (43.8%) | 0.18 |
| OS | 324 (54.3%) | 97 (58.8%) | 127 (50.0%) | 100 (56.2%) |
Notes: * indicates statistical significance (p<0.05). †Ethnicity data were unavailable for 2 individuals in the low-risk RD-OGI group. Percentages were calculated for the overall cohort and within each risk level, and are presented to total 100% for each characteristic listed in the leftmost column.
Statistical Analysis
To compare patient demographic and clinical characteristics across RD risk groups, means and standard deviations or median and interquartile ranges are presented for continuous variables. Frequencies and percentages are presented for categorical variables. P values were calculated using ANOVA or Kruskal–Wallis tests for continuous variables and chi-square or Fisher’s exact tests for categorical variables. To examine how well the RD-OGI score predicts RD status at different follow-up periods, area under the receiver operating characteristic (ROC) curve and its confidence intervals were calculated using nonparametric ROC analysis. An a priori precision assessment for the ROC analysis was conducted. Using the number of eyes with follow-up and the number of RDs at each follow-up interval, all available sample sizes yielded two-sided 95% confidence intervals for the AUC with widths below 0.11 when assuming a true AUC of 0.75. This level of precision is consistent with accepted standards for diagnostic accuracy studies. Complete-case analysis was applied throughout because the extent of missing data in this study was minimal with only 3.6% missing OGI scores. Given the very low proportion of missing observations, multiple imputation was not expected to meaningfully change the results or improve precision. For this reason, and to maintain transparency and interpretability, the complete-case results were presented as the primary analysis. Kaplan–Meier curves for the overall cohort and by risk groups were plotted for patients’ status at one-year follow-up as well as over the entire study period. All analyses were carried out in the statistical software Stata v18.0. A p value less than 0.05 is considered statistically significant.
Results
Initially, 663 patients presenting to the emergency department with acute OGI between 2006 and 2017 were identified. After exclusion of 24 individuals with insufficient data to calculate RD-OGI score and then 42 individuals with less than 30 days of follow up, a total of 597 eyes with OGI were included in the final analyses (Figure 1). Of these, 165 (27.6%) were classified as low-risk, 254 (42.5%) as moderate-risk, and 178 (29.8%) as high-risk. The population was predominantly male (n=466, 78.1%), with a similar distribution in the laterality of the affected eye (p=0.18). The mean age was significantly different between the groups (47.4±24.2 years, 40.2±21.4 years and 34.4±17.7 years, for the high, moderate and low risk groups, respectively; p<0.001). The demographics of the participants included in the analyses stratified by RD-OGI risk are presented in Table 1.
Figure 1.
Flow diagram of inclusion and exclusion criteria.
VA at presentation was significantly different among all groups (p<0.001). In the low-risk group, the majority of the patients presented with a VA of ≥20/40 (n=57, 34.5%), while none of the eyes presented with light perception (LP) or no light perception (NLP) vision. In the moderate-risk group, the majority presented with hand motion (HM; n=116, 45.7%), followed by LP (n=78, 30.7%) and counting fingers (CF; n=27, 10.6%). In the high-risk group, all eyes had VA of CF (n=14, 7.9%) or worse, including HM (n=35, 19.7%), LP (n=84, 47.2%) and NLP (n=45, 25.3%).
Zone of injury was reported according to the zones involved at initial presentation. Highest zone of injury was zone I among all cases (255 eyes, 42.7%), followed by zone II (n=214, 35.8%) and zone III (n=128, 21.4%). In the low-risk group, most eyes presented with Zone I injuries (n=140, 84.8%), followed by Zone II (n=51, 30.9%) and Zone III (n=16, 9.7%). Similarly, Zone I injuries were most common in the moderate-risk group (n=190, 74.8%), with fewer cases involving Zone II (n=100, 39.4%) and Zone III (n=47, 18.6%). In the high-risk group, the majority of injuries were categorized as Zone II (n=155, 87.1%), followed by Zone I (n=98, 55.1%) and Zone III (n=65, 36.5%).
Less than half of all eyes (n=240, 40.2%) presented with VH, with only 2 eyes (1.2%) in the low-risk group, compared with 74 eyes (29.1%) in the moderate-risk group and 164 eyes (92.1%) in the high-risk group. As expected, the presence of VH was significantly higher as the risk for RD increased (p<0.001). The distribution of VA, ZOI, and presence of VH across the risk groups is summarized in Table 2.
Table 2.
Distribution of Visual Acuity (VA), Zone of Injury (ZOI), Presence of Vitreous Hemorrhage (VH) and Time Between Initial Repair to Last Follow Up Stratified by Low Risk, Moderate Risk and High Risk RD-OGI Score
| All (n=597) | Low-Risk (n=165) |
Moderate-Risk (n=254) |
High-Risk (n=178) |
p-value | ||
|---|---|---|---|---|---|---|
| VA at presentation | ≥20/40 | 65 (10.9%) | 57 (34.5%) | 8 (3.1%) | 0 (0.0%) | <0.001* |
| 20/40- 20/70 | 35 (5.9%) | 31 (18.8%) | 4 (1.6%) | 0 (0.0%) | ||
| 20/80-20/200 | 52 (8.7%) | 40 (24.2%) | 12 (4.7%) | 0 (0.0%) | ||
| 20/200-20/400 | 15 (2.5%) | 12 (7.3%) | 3 (1.2%) | 0 (0.0%) | ||
| CF | 64 (10.7%) | 23 (13.9%) | 27 (10.6%) | 14 (7.9%) | ||
| HM | 153 (25.6%) | 2 (1.2%) | 116 (45.7%) | 35 (19.7%) | ||
| LP | 162 (27.1%) | 0 (0.0%) | 78 (30.7%) | 84 (47.2%) | ||
| NLP | 51 (8.5%) | 0 (0.0%) | 6 (2.4%) | 45 (25.3%) | ||
| Zone I Injury | No | 169 (28.3%) | 25 (15.2%) | 64 (25.2%) | 80 (44.9%) | <0.001* |
| Yes | 428 (71.7%) | 140 (84.8%) | 190 (74.8%) | 98 (55.1%) | ||
| Zone II Injury | No | 291 (48.7%) | 114 (69.1%) | 154 (60.6%) | 23 (12.9%) | <0.001* |
| Yes | 306 (51.3%) | 51 (30.9%) | 100 (39.4%) | 155 (87.1%) | ||
| Zone III Injury | No | 468 (78.5%) | 149 (90.3%) | 206 (81.4%) | 113 (63.5%) | <0.001* |
| Yes | 128 (21.5%) | 16 (9.7%) | 47 (18.6%) | 65 (36.5%) | ||
| Highest Zone of Injury | Zone I | 255 (42.7%) | 106 (64.2%) | 130 (51.2%) | 19 (10.7%) | <0.001* |
| Zone II | 214 (35.8%) | 43 (26.1%) | 77 (30.3%) | 94 (52.8%) | ||
| Zone III | 128 (21.4%) | 16 (9.7%) | 47 (18.5%) | 65 (36.5%) | ||
| Vitreous Hemorrhage | No | 357 (59.8%) | 163 (98.8%) | 180 (70.9%) | 14 (7.9%) | <0.001* |
| Yes | 240 (40.2%) | 2 (1.2%) | 74 (29.1%) | 164 (92.1%) | ||
| Initial repair to last follow-up (days), median (IQR) | 405.0 (101.0, 1220.0) | 298.0 (93.0, 924.0) | 476.0 (113.0, 1240.0) | 394.0 (97.0, 1428.0) | 0.53 |
Notes: * indicates statistical significance (p<0.05). Percentages were calculated for the overall cohort and within each risk level, and are presented to total 100% for each characteristic listed in the leftmost column. Zone of injury categories are not mutually exclusive (ie, an eye with Zones I and II injury would be classified as “Yes” for both Zone 1 and Zone 2 separately, but only “Zone 2” for Highest ZOI).
Abbreviations: CF, counting fingers; HM, hand motion; LP, light perception; NLP, no light perception.
A total of 144 (39.2%) eyes were determined to develop RD by 365 days and 150 (49.0%) by the end of the study at 720 days. The distribution of patients according to the risk of RD development within 30, 90, 180, 365, and 720 days is presented in Table 3. At one month after surgical repair, RD was observed in 6 eyes (6.1%) in the low-risk group, 36 eyes (36.4%) in the moderate-risk group, and 57 eyes (57.6%) in the high-risk group. We observed that majority of the RDs developed within the first 6 months of follow up. By the six-month mark, the number of eyes developing RD had increased to 8 in the low-risk group (5.6%), 55 in the moderate-risk group (38.5%), and 80 in the high-risk group (55.9%), with a total of 143 eyes (33.7%). After six months, the development of RD became significantly less frequent. At the one year follow up, only one new RD case in the low risk group was reported, increasing the total to 9 eyes (6.3%) in this group. There were no new RD cases observed in the moderate or high risk groups. By the two-year follow-up, only 6 additional cases with RD were identified, with no new cases in the low-risk group, three in the moderate-risk group (n=58, 38.7%), and three in the high-risk group (n=83, 55.3%).
Table 3.
RD-OGI Score Among Eyes Developing RD
| RD-OGI Score | Length of Follow-Up Period | |||||
|---|---|---|---|---|---|---|
| 30 Days, N=549 |
90 Days, N=485 |
180 Days, N=424 |
365 Days, N=367 |
720 Days, N=306 |
||
| Total no. of eyes with RD within timeframe, N (%) | 99 (18.0%) | 128 (26.4%) | 143 (33.7%) | 144 (39.2%) | 150 (49.0%) | |
| Low-risk (0–2) | 0.0 | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) |
| 0.5 | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | |
| 1.5 | 3 (3.0%) | 3 (2.3%) | 4 (2.8%) | 4 (2.8%) | 4 (2.7%) | |
| 2.0 | 3 (3.0%) | 4 (3.1%) | 4 (2.8%) | 5 (3.5%) | 5 (3.3%) | |
| Total low-risk | 6 (6.1%) | 7 (5.5%) | 8 (5.6%) | 9 (6.3%) | 9 (6.0%) | |
| Moderate-risk (2.5–4.5) | 2.5 | 3 (3.0%) | 5 (3.9%) | 7 (4.9%) | 7 (4.9%) | 7 (4.7%) |
| 3.0 | 5 (5.1%) | 7 (5.5%) | 10 (7.0%) | 10 (6.9%) | 12 (8.0%) | |
| 3.5 | 0 (0.0%) | 1 (0.8%) | 1 (0.7%) | 1 (0.7%) | 1 (0.7%) | |
| 4.0 | 11 (11.1%) | 16 (12.5%) | 18 (12.6%) | 18 (12.5%) | 19 (12.7%) | |
| 4.5 | 17 (17.2%) | 18 (14.1%) | 19 (13.3%) | 19 (13.2%) | 19 (12.7%) | |
| Total moderate-risk | 36 (36.4%) | 47 (36.7%) | 55 (38.5%) | 55 (38.2%) | 58 (38.7%) | |
| High-risk (5–7.5) | 5.0 | 9 (9.1%) | 11 (8.6%) | 12 (8.4%) | 12 (8.3%) | 12 (8.0%) |
| 5.5 | 4 (4.0%) | 5 (3.9%) | 6 (4.2%) | 6 (4.2%) | 6 (4.0%) | |
| 6.0 | 12 (12.1%) | 18 (14.1%) | 19 (13.3%) | 19 (13.2%) | 20 (13.3%) | |
| 6.5 | 25 (25.3%) | 30 (23.4%) | 33 (23.1%) | 33 (22.9%) | 35 (23.3%) | |
| 7.5 | 7 (7.1%) | 10 (7.8%) | 10 (7.0%) | 10 (6.9%) | 10 (6.7%) | |
| Total high-risk | 57 (57.6%) | 74 (57.8%) | 80 (55.9%) | 80 (55.6%) | 83 (55.3%) | |
Notes: Percentages within low-, moderate-, and high-risk groupings represent proportion of total eyes with RD within the specified timeframe.
The overall Kaplan–Meier survival curve highlights a significant incidence of RD within the first six months following globe repair, with a plateau thereafter (Figure 2a). Additionally, we observed that the RD-OGI score exhibited its strongest predictive performance at 6 months, with an area under the curve (AUC) of 0.774 (95% confidence interval [CI]: 0.730–0.819). AUC values for 30, 90, 180, 365 and 720 days can be found in Table 4. Stratified curves show that the high-risk group experiences a steep and rapid decline in RD-free survival, with most cases occurring early during the follow up period (within 6 months). The moderate-risk group shows a gradual decline over time, while the low-risk group remains relatively stable, with minimal RD events throughout the first year follow-up period (Figure 2b). Similarly, the survival curves present a similar trend for the final follow up at 720 days with most RD cases occurring in the first 6 months both overall and after stratification into groups, with a plateau thereafter (Figure 2c and d).
Figure 2.
Kaplan Meier curves for retinal detachment risk within 365 days; overall cohort (a) and stratified by risk groups (b). Kaplan Meier curves for retinal detachment risk at final follow up; overall cohort (c) and stratified by risk groups (d).
Table 4.
Predictive Performance of RD-OGI Score at Different Intervals Following Injury
| Follow-Up Time | Number of Eyes with Follow-Up |
ROC Area | 95% Confidence Interval |
|---|---|---|---|
| 30 Days | 549 | 0.756 | 0.709–0.802 |
| 90 Days | 485 | 0.771 | 0.728–0.815 |
| 180 Days | 424 | 0.774 | 0.730–0.819 |
| 365 Days | 367 | 0.751 | 0.702–0.801 |
| 720 Days | 306 | 0.739 | 0.684–0.794 |
Discussion
In this large retrospective analysis of OGIs presenting to a single tertiary-care trauma referral center, we observed an overall moderate performance of the RD-OGI score to predict RD development at 6 months up to 720 days post-injury. The Ocular Trauma Classification Group categorized ocular injury based primarily on the type and grade of injury (determined by VA at presentation), the presence of a relative afferent pupillary defect (RAPD), and the ZOI at presentation.15 The more recently described RD-OGI score was developed to bridge the gap between the initial presentation and outcomes of OGI by offering a validated and reproducible estimate of a patient’s risk of developing RD, one of the most common complications impacting visual outcomes after OGI.10,11
While our results are similar to the initial reports of the RD-OGI scoring system for the low and moderate risk groups, we observed a lower observed frequency of RD in the high-risk group. In our cohort, of the 144 eyes that had developed RD by the 12 month follow up visit, 9 (6.3%) were in the low risk, 55 (38.2%) were in the moderate risk and 80 (55.6%) in the high risk group. In comparison, by the one year follow up, the RD-OGI derivation cohort showed RD rates of 3%, 29%, and 73%, while their validation cohort had rates of 0%, 35%, and 86% for the low-, moderate-, and high-risk groups, respectively.10,11
The lower frequency of RD in our high risk group could be related to the lack of universal use of ancillary diagnostic tests to confirm the diagnosis. We observed that 8% (n=11) of the high-risk eyes diagnosed with no RD in our cohort did not undergo B scan or CT imaging, possibly as these cases had little to no potential for visual improvement, which meant that the presence or absence of RD could not be conclusively determined. As no definitive RD diagnosis was made in these cases, they were all considered in the “no RD” group in our analyses. In the initial report in which the RD-OGI score was derived, patients with an incomplete workup were specifically excluded. In a recent study by Perez et al, the authors encountered a similar situation, where many of the high-risk patients did not undergo detailed retinal evaluation.12
Regarding prediction of RD risk, the RD-OGI score demonstrated excellent performance at 30 days in the initial published validation cohort, with an AUC of 0.939, indicating strong discriminative ability.11 Similarly, Perez et al also found an excellent discriminatory ability of the RD-OGI score for predicting RD incidence with an AUC of 0.84 (95% CI, 0.79–0.89) at 30 days, still higher than the results observed in our study at this timepoint.12 Our study recorded an AUC of 0.756 at 30 days, reflecting only moderate predictive performance. Additionally, while the initial validation study consistently reported high AUC values throughout the follow-up period (all above 0.90) until 365 days, our study did not observe any AUC values exceeding 0.8. The highest AUC observed in our cohort was achieved at 180 days (AUC= 0.774), representing relatively better predictive accuracy at 6 months follow up, while only achieving an AUC of 0.739 in the last follow up at 2 years. There are several possible reasons for lower predictive performance of the RD-OGI score in our study.
First, these differences could be due in part to undiagnosed RD cases in our cohort since, as noted above, imaging with B scan ultrasound in the setting of media opacity was not uniformly performed. Second, there could be variations in study populations, including specific mechanisms of OGI, highlighting the need for further research to clarify the RD-OGI score’s applicability across diverse clinical settings. Finally, as our study included a longer follow-up period, the low incidence of RD at later timepoints following OGI may have influenced overall predictive performance. The initial published validation study reported the longest time to RD as 133 days.11
The lower predictive performance of the RD-OGI score in our cohort raises the possibility that incorporation of additional clinical features at presentation beyond VA, ZOI, and VH may be helpful to better predict risk in some populations. A derivation cohort from our institution in the future, for example, could theoretically assess whether inclusion of findings such as an RAPD, presence of a hyphema, or additional imaging biomarkers could improve an RD-OGI score rubric tailored to the specific population. Interestingly, recent publications have suggested that artificial intelligence-based imaging analyses may provide utility in prediction of RD recurrence following initial repair but whether these could be applied to help predict RD risk in the setting of trauma when visualization of the fundus can be limited remains unclear.16
While providing insights into the generalizability and long-term predictive value of the RD-OGI score, our study has several limitations. First, its retrospective nature and the lack of universal use of imaging to confirm presence or absence of RD may have underestimated our presumed number of RDs in eyes with initial media opacity limiting view to the fundus. Second, we only report data from a single tertiary care center albeit with a high number of eyes included in the study. Third, improvements or changes in standard of care over the extended study period may have influenced the acquired data and contributed to variability in management and outcomes. Fourth, although all patients were treated at a single academic institution, variation in attending supervision and the involvement of trainees may have affected initial clinical assessments. As first-line evaluations were often performed by rotating trainees, the quality and completeness of data collected at the initial presentation may have been subject to interobserver variability. Finally, the lack of analysis of secondary surgical interventions, such as repeat RD repair and management of elevated intraocular pressure, may have affected our 2-year outcomes and contributed to less robust results.
Conclusion
Overall, our study found that the RD-OGI score demonstrated moderate predictive performance for RD risk at 6 months (AUC 0.774), which remained similar by 24 months (AUC 0.739). These findings suggest that, while the RD-OGI score provides a valuable tool for assessing risk of RD following OGI and can guide providers in counseling high-risk patients for timely vitreoretinal evaluation, the strength of its predictive value may vary depending on real-world factors related to clinical setting or patient population. When considering application of the RD-OGI score to longer-term follow-up periods, refinement of the model could be considered to optimize predictive performance.
Data Sharing Statement
Deidentified data are available upon request to the corresponding author.
Ethics Approval and Informed Consent
The study was conducted in accordance with the tenets of the Declaration of Helsinki and approved by the Johns Hopkins Medicine Institutional Review Board (IRB00132759). Due to the retrospective nature of the study and minimal risk involved to the privacy of individual participants, a waiver for written informed consent was approved by the Johns Hopkins Medicine IRB.
Disclosure
Ms Jiangxia Wang reports Support for the manuscript from National Center for Advancing Translational Sciences (NCATS), during the conduct of the study. Dr Fasika Woreta reports Honoraria from Walter Reed, Vidico Medical Education; Leadership or fiduciary roles from President, outside the submitted work. Dr TY Liu reports Support for the manuscript from Gills AI Innovation Center, during the conduct of the study; Grants or contracts from Research to Prevent Blindness Career Development Award, outside the submitted work. The authors report no other conflicts of interest in this work.
References
- 1.Mir TA, Canner JK, Zafar S, Srikumaran D, Friedman DS, Woreta FA. Characteristics of open globe injuries in the United States from 2006 to 2014. JAMA Ophthalmol. 2020;138(3):268. doi: 10.1001/jamaophthalmol.2019.5823 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Zhou Y, DiSclafani M, Jeang L, Shah AA. Open globe injuries: review of evaluation, management, and surgical pearls. OPTH. 2022;16:2545–9. doi: 10.2147/OPTH.S372011 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Mayer CS, Reznicek L, Baur ID, Khoramnia R. Open globe injuries: classifications and prognostic factors for functional outcome. Diagnostics. 2021;11(10):1851. doi: 10.3390/diagnostics11101851 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Vought R, Vought V, Henry RK, Oydanich M, Khouri AS. Outpatient visits and secondary surgeries following open globe injuries: a single institution retrospective analysis. Int Ophthalmol. 2026;46(1):112. doi: 10.1007/s10792-026-03991-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Huong PTT, Anh BTV, et al; Hanoi Medical University, Hanoi 100000, Vietnam; Vietnam National Eye Hospital, Hanoi 100000, Vietnam. Visual prognosis following open globe injury: clinical characteristics and risk factors from Vietnam National Eye Hospital. Int J Ophthalmol. 2026;19(2):346–353. doi: 10.18240/ijo.2026.02.18 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Dave VP, Das AV, Nayak S, Panchal B, Kuhn F. Clinical presentations, management, and factors affecting outcomes in posterior segment open globe injuries: an analysis of 2,360 eyes. Ophthalmologica. 2023;246(2):150–157. doi: 10.1159/000529680 [DOI] [PubMed] [Google Scholar]
- 7.Sheth N, Chang AY, Bharadwaj A, Wu D, Ivakhnitskaia E, Mieler WF. Early retinal detachments after open globe repair: incidence and risk factors. Retina. 2025;45(4):614–620. doi: 10.1097/IAE.0000000000004348 [DOI] [PubMed] [Google Scholar]
- 8.Sinha AK, Durrani AF, Li KX, et al. Retinal detachments after open-globe injury. Ophthalmol Retina. 2024;8(4):340–349. doi: 10.1016/j.oret.2023.10.008 [DOI] [PubMed] [Google Scholar]
- 9.Sheth N, Chang AY, Bryan JM, Massengill MT, Lim JI. Outcomes of open-globe injuries with associated retinal detachment: experience at an ocular trauma center. J VitreoRet Dis. 2025;9(2):180–186. doi: 10.1177/24741264241301763 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Stryjewski TP, Andreoli CM, Eliott D. Retinal Detachment after Open Globe Injury. Ophthalmology. 2014;121(1):327–333. doi: 10.1016/j.ophtha.2013.06.045 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Brodowska K, Stryjewski TP, Papavasileiou E, Chee YE, Eliott D. Validation of the Retinal Detachment after Open Globe Injury (RD-OGI) score as an effective tool for predicting retinal detachment. Ophthalmology. 2017;124(5):674–678. doi: 10.1016/j.ophtha.2016.12.032 [DOI] [PubMed] [Google Scholar]
- 12.Perez EA, Scott NL, Russell JF. Outcomes of retinal detachment after open-globe injury, and independent validation of the retinal detachment after open-globe injury scoring system. Ophthalmology. 2024;131(4):412–421. doi: 10.1016/j.ophtha.2023.10.030 [DOI] [PubMed] [Google Scholar]
- 13.Cardillo JA, Stout JT, LaBree L, et al. Post-traumatic proliferative vitreoretinopathy. Ophthalmology. 1997;104(7):1166–1173. doi: 10.1016/S0161-6420(97)30167-5 [DOI] [PubMed] [Google Scholar]
- 14.Mura M, Iannetta D, Pellegrini M, et al. Long-term functional and structural outcomes after large chorioretinectomy for ruptured globe following blunt trauma. Int J Retin Vitr. 2023;9(1):52. doi: 10.1186/s40942-023-00492-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Pieramici DJ, Sternberg P, Aaberg TM, et al. A system for classifying mechanical injuries of the eye (Globe). Am J Ophthalmol. 1997;123(6):820–831. doi: 10.1016/S0002-9394(14)71132-8 [DOI] [PubMed] [Google Scholar]
- 16.Catania F, Chapron T, Crincoli E, et al. Deep Learning for prediction of late recurrence of retinal detachment using preoperative and postoperative ultra-wide field imaging. Acta Ophthalmologica. 2024;102(7). doi: 10.1111/aos.16693 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Deidentified data are available upon request to the corresponding author.