Trends in Retinopathy of Prematurity Screening and Treatment: 2008–2018

Vermont Oxford Network (VON) is a nonprofit, voluntary worldwide collaboration dedicated to improving the quality, safety, and value of neonatal intensive care.⁷  VON member hospitals contributed standardized data on very low birth weight (VLBW) infants, defined here as those with a GA of 22 to 29 weeks or BW of 401 to 1500 g who were inborn or transferred to the hospital within 28 days of birth.⁸  GA was determined by the best estimate using the following hierarchy: obstetrical measures based on last menstrual period, obstetrical parameters, then prenatal ultrasound, followed by neonatologist’s estimate based on physical criteria and examination.⁹ 

We included data for infants at US hospitals born from January 1, 2008, to December 31, 2018. Supplemental Table 4 reveals the list of hospitals included in the analysis. For this study, we excluded the following: infants with implausible BWs (defined as >3 standard deviations from the mean for age and sex),¹⁰  infants missing data on the ophthalmologic examination or length of hospital stay, and infants who were transferred to a non-VON hospital, were discharged, or died before their first recommended ROP screening examination at 4 weeks chronological age or 31 weeks’ postmenstrual age (PMA), whichever was later⁶  (Fig 1).

The American Academy of Pediatrics policy statement “Screening Examination of Premature Infants for ROP” recommends that screening commence at 4 weeks’ chronological age or 31 weeks’ PMA, whichever is later.⁶  ROP screening was defined as any indirect ophthalmoscopy examination for ROP performed at any time until hospital discharge or transfer to a non-VON hospital among infants who were hospitalized at the recommended screening age on the basis of the American Academy of Pediatrics statement. Among infants who received an eye examination, ROP severity was defined as the worst stage of ROP documented on any examination in the eye with the most advanced stage. ROP stages (from least to most severe) include the following: stage 0 = immature retinal vascularization, no ROP; stage 1 = presence of a demarcation line between vascular and avascular retina; stage 2 = presence of an elevated ridge at the vascular or avascular junction; stage 3 = extraretinal fibrovascular proliferation extending from the ridge; stage 4 = partial retinal detachment; and stage 5 = total retinal detachment.¹¹  For this study, any ROP was defined as stages 1 to 5 and severe ROP was defined as stages 3 to 5. Retinal ablation was defined as retinal cryotherapy and/or laser ablation.

VON member hospitals completed surveys of center characteristics annually. By using responses to whether the center was required by state regulation to transfer infants to another center for assisted ventilation on the basis of infant characteristics or duration of ventilation required or whether at least 1 of 8 surgeries was performed at the center (omphalocele repair, ventriculoperitoneal shunt, tracheoesophageal fistula or esophageal atresia repair, bowel resection or reanastomosis, meningomyelocele repair, patent ductus arteriosus ligation, cardiac catheterization, or cardiac surgery requiring bypass), centers were divided into 3 groups: type A has ventilation restrictions or performs none of the surgeries named above; type B has no ventilation restrictions and performs at least 1 of the surgeries named above, not including cardiac surgery requiring bypass; and type C has no ventilation restrictions and performs cardiac surgery requiring bypass. Infants were categorized according to the NICU type from which they were discharged.

All data analyses were performed by using R (v.3.5.3; R Foundation for Statistical Computing, Vienna, Austria). We calculated rates of ROP screening and treatment and the incidence and severity of ROP from 2008 to 2018. Data on treatment of ROP with anti-VEGF injection were available beginning in 2012. We evaluated trends over time using logistic regression with birth year as a linear effect. We repeated the trend analysis, adjusting for the infant characteristics of GA, sex, multiple gestation, small for gestational age (SGA), and antenatal steroid exposure. SGA was defined within categories of sex, race, ethnicity, and multiple birth as BW below the 10th percentile on the basis of smoothed curves constructed by using the US natality dataset.¹²  Infants were considered to have exposure to antenatal steroids if betamethasone, dexamethasone, or hydrocortisone were administered intramuscularly or intravenously to the mother during pregnancy at any time before delivery.⁹ 

Because VON membership changed over the study period, we performed a sensitivity analysis repeating the above analyses, including only hospitals that participated all 11 years.

The University of Vermont Institutional Review Board reviewed this proposal and determined it was not human subjects research.

Of all US VLBW infants in the VON database from 2008 to 2018 (n = 491 842), this analysis includes standardized data collected on 381 065 infants (77%) who were still hospitalized at the recommended age for ROP screening at 819 US hospitals (Fig 1). There were 570 hospitals included in 2008, which increased to 730 hospitals in 2018.

In our sample, the median BW was 1090 g, the median GA was 28 + 5/7 weeks, 51% of infants were boys, 73% of gestations were singletons, 18% of infants were SGA, and maternal race and/or ethnicity (classified by interviewing the mother or review of the birth certificate or medical record, in that order of preference) was reported as 55% white, 30% Black or African American, and 18% Hispanic (Table 1).

Over the time period, the rate of ROP screening among infants still hospitalized at their recommended screening ages increased from 89% in 2008 to 91% in 2018 (trend P < .001) (Table 2). Among screened infants, overall incidence of any ROP (ie, stages 1–5) decreased from 37% in 2008 to 32% in 2018 (trend P < .001), and severe ROP (ie, stages 3–5) decreased from 8% in 2008 to 6% in 2018 (trend P < .001) (Table 2, Fig 2A). Among infants with a BW <1251 g, the overall incidence of any ROP was 44%, and, among those with a BW <1500 g, the overall incidence of any ROP was 35%. Rates of treatment with either retinal ablation or anti-VEGF injection decreased from 6% in 2008 to 4% in 2018 (trend P < .001) (Table 2). Whereas rates of retinal ablation decreased from 6% in 2008 to 2% in 2018 (trend P < .001), rates of anti-VEGF injections increased from 1% in 2012 to 2% in 2018 (trend P < .001) (Table 2, Fig 2B). Rates of treatment with both retinal ablation and anti-VEGF injection on the same infant also increased from 0.3% in 2012 to 0.6% in 2018 (trend P < .001). Each of the trends remained statistically significant over the time period after adjusting for infant characteristics (ie, GA, sex, multiple gestation, SGA, and antenatal steroid exposure).

Of the 381 065 infants in our study, 36 403 (10%) were not screened for ROP over the time period (Table 3). Of infants who were in the hospital at the recommended screening age but were not screened for ROP, 51% had a GA ≤30 completed weeks, 98% had a BW ≤1500 g, 86% were discharged home from the hospital, 10% were transferred to a non-VON hospital, and 5% died before discharge. The median PMA at discharge for infants who were not screened was 36 + 5/7 weeks (interquartile range: 35 + 4/7 to 38 weeks) and the median length of hospital stay for infants not screened was 37 days (interquartile range: 31–50 days). Of the 381 065 infants in the study, those who were in the hospital at the recommended screening age but were not screened for ROP included 10 725 infants (15%) at type A compared with 17 197 (9%) at type B and 8470 (7%) at type C centers (Fig 2C).

The common set of hospitals that participated in VON for all 11 years included 504 hospitals. Results using the common set of hospitals did not differ from results using the full US database (results not reported).

Among hospitals in the United States participating in VON from 2008 to 2018, more infants were screened for ROP, less overall and severe ROP were reported, and, although less retinal ablation was performed, the use of anti-VEGF treatment has increased. Infants not screened for ROP tended to have higher GAs and BWs and were more likely to be at lower acuity (type A) NICUs.

Although our findings were encouraging and reveal an improvement in care regarding a higher percentage of infants being screened for ROP according to national guidelines,⁶  over the time period, 10% of the infants who were in the hospital at the recommended ages for screening were still not screened for ROP. It is important to figure out why not all infants are being screened for ROP while in the hospital. How are infants being identified for ROP screening examinations, and is this being performed in a timely manner? Are infants being missed because of lack of personnel, oversight, or understanding of current screening guidelines? One study reported that, in a survey of neonatologists, only 19% used the currently recommended GA and 86% the BW criterion.¹³  We found that infants who did not receive at least 1 ROP screening examination while hospitalized at the recommended screening age had a higher median GA and BW than those receiving screening examinations. Of those infants in our study not screened for ROP, 49% met screening criteria by BW only (ie, had a GA >30 weeks and BW ≤1500 g) and 2% met screening criteria by GA only (ie, had a GA <30 weeks and BW >1500 g). Also, we found that the lower acuity (type A) NICUs were almost twice as likely as the higher acuity (types B and C) NICUs to not have performed at least 1 ROP screening examination. Although it is possible that the infants missed were too systemically unstable to be examined in the hospital, only 5% of infants died before discharge. Additionally, because we examined “any ROP” screening in the hospital rather than serial and timely ROP screening, we may be underestimating the percentage of infants adequately screened for ROP in the hospital. Elucidating the causes for these missed opportunities could allow us the ability to improve health care delivery and address them on an individual, institutional, or national level.

From 2008 to 2018, there were significant decreases in the incidence of any (stages 1–5) reported ROP and severe (stages 3–5) ROP. The declines in any reported ROP and severe ROP are most likely due to improvements in perinatal and neonatal care and practices.¹⁴  Among infants with BW <1251 g, the incidence of any ROP has been reported to be 65.8% from January 1986 to November 1987 among 23 participating study centers in the Cryotherapy for ROP study,¹⁵  68% from October 2000 to September 2002 at 26 participating study centers in the Early Treatment for ROP study,¹⁶  and 63.7% from May 2011 to October 2013 at 13 North American (12 in the United States and 1 in Canada) centers in the Telemedicine Approaches for the Evaluation of Acute-Phase ROP study.^17,18  Using a comprehensive database from New York State, with initial length of hospital stay >28 days and date of discharge from January 1996 to December 2000, the incidence of any ROP among infants with BW <1500 g was 20.3%.¹⁹  Using a national US database from 1997 to 2005 among all newborns with length of stay >28 days, the incidence of any ROP was 15.58% among all newborns.²⁰  Using a largely representative national neonatal database from 2008 to 2018 from 819 NICUs across the United States, we found that, among infants with BW <1251 g, the incidence of any ROP was 44%, and, among infants with BW<1500 g, the incidence of any ROP was 35%. The differences in reported incidences across studies are likely due to differences in study design or methodology and study population. The Cryotherapy for ROP, Early Treatment for ROP, and Telemedicine Approaches for the Evaluation of Acute-Phase ROP studies were prospective, multicenter studies that included centers with higher-risk infants (ie, primarily academic medical centers) and only included infants with BW <1251 g, thus reporting the highest incidences of ROP. The studies based on state and national databases were more comprehensive and, like our study, included centers with lower-risk infants (eg, type A NICUs), which explains lower incidences of ROP. Because of these differences, we focused on evaluating trends of ROP incidence over time using data collected over time by an established neonatal network.

From 2008 to 2018, there were significant decreases in any treatment of ROP and retinal ablation for ROP and a significant increase in intravitreal anti-VEGF therapy. The decreasing trend in any treatment of ROP (ie, any retinal ablation or anti-VEGF injection) coincides with the decreasing trend of severe ROP. It is not surprising that the rates of intravitreal injection for the treatment of ROP have increased since the 2011 publication reporting that intravitreal bevacizumab monotherapy had a significant benefit over conventional laser ablation for the treatment of a subgroup of type 1 (treatment-indicated) ROP.⁵  In our data, there was also a significant increase in the use of both retinal ablation and anti-VEGF injection on the same infant. The use of anti-VEGF treatment is sometimes followed by laser ablation for either recurrent ROP or for prophylaxis (ie, performing laser ablation to any remaining avascular retina to prevent ROP recurrence).²¹  Because anti-VEGF treatments were only recently added to the ophthalmologist’s armamentarium of treatments for ROP, there are currently no standardized guidelines for whether laser ablation should be performed prophylactically before an infant’s discharge from the hospital. In some hospitals, ophthalmologists perform laser ablation to any significant avascular retina before discharge, whereas in other hospitals they do not. Infants who received an anti-VEGF treatment followed by prophylactic laser ablation may inflate our reported rates of treatment of ROP, specifically overestimating the rates of laser ablation for presumed recurrent or severe ROP. Thus, it is likely that the rate of laser ablation for recurrent or severe ROP is declining at a faster rate than our results imply.

This study affords us the unique opportunity to evaluate the trend of ROP screening, incidence, severity, and treatment practices in US hospitals from 2008 to 2018 but should be taken in light of some limitations. Because VON’s VLBW database includes infants with GA <30 weeks or BW ≤1500 g, the subgroup of infants with a GA ≥30 weeks and BW >1500 g was not included in this study. Additionally, we do not have ROP information on infants who were not screened, who tended to have a higher GA and BW than screened infants. By not including these higher GA infants who likely have a lower incidence of ROP, less severe ROP, and lower need for treatment compared with lower GA infants, we are likely overestimating the prevalence of ROP, severe ROP, and treatment among infants in the hospital who should be screened for ROP. Another limitation is that our database only captures data while the infant is in a VON member hospital, so we do not include ROP occurrence or treatment after discharge or that occurs at a non-VON hospital. We excluded infants who were discharged or transferred to a non-VON hospital before the date of recommended screening. However, there may be some infants still hospitalized at the date of recommended screening who receive their first ROP screening examination as an outpatient. Although this would underestimate screening rates, this also reflects delayed screening that does not comply with US screening guidelines recommending the first screening examination at 4 weeks’ chronological age or 31 weeks’ PMA, whichever comes later.⁶  In addition, we do not collect data confirming that those treated met treatment criteria (ie, zone and stage of ROP and the presence of plus disease at the time of treatment) or on visual outcome. Because of the above limitations, we do not report a true incidence of ROP severity or treatment and we cannot report if visual outcomes have changed over time. The goal and strength of this study is its ability to evaluate trends of ROP screening, incidence, severity, and treatment practices over time using an established neonatal network that is dedicated to improving the quality, safety, and value of neonatal intensive care.

Our findings provide insight into how hospitals across the country are performing and their outcomes over the last decade regarding ROP (ie, the National Quality Forum has endorsed “the proportion of infants 22–29 weeks gestation screened for ROP” as a quality measure on perinatal care). Although we show overall improvement in ROP screening in US hospitals, we found that there is room for improvement because 10% of infants were not screened for ROP while in the hospital. Our findings suggest that opportunities for improving rates of screening include education on the current screening guidelines (in particular at what GAs and BWs infants should be screened) and focusing our efforts on increasing ROP screening in lower acuity NICUs. Further studies evaluating these missed opportunities will be important to allow us to continue to improve our care for these infants at risk and further reduce the risk of preventable childhood blindness due to ROP.

We thank our colleagues who submit data to VON on behalf of infants and their families, allowing us the unique opportunity to evaluate this valuable information to understand the performance of US hospitals over time and the opportunity for quality assessment and improvement to ultimately help care for other infants at risk for ROP. A list of hospitals included in this analysis is in Supplemental Table 4.

anti-VEGF

anti–vascular endothelial growth factor

BW

birth weight

GA

gestational age

MA

postmenstrual age

ROP

retinopathy of prematurity

SGA

small for gestational age

VLBW

very low birth weight

VON

Vermont Oxford Network