This cohort study examines data for preterm neonates to investigate the association of socioeconomic factors in the context of race and ethnicity with retinopathy of prematurity outcomes.
Key Points
Question
How are socioeconomic factors, in the context of race and ethnicity, associated with retinopathy of prematurity (ROP) outcomes?
Findings
In this cohort study including 1234 preterm neonates, Hispanic infants were more likely to develop ROP and had more severe ROP. Associations between race and ethnicity and ROP outcomes were mainly driven by disparities in gestational age, which in our data set were mainly explained by neighborhood income level.
Meaning
When placed in the context of socioeconomic factors and gestational age, race and ethnicity were not associated with ROP diagnosis; research on racial and ethnic disparities in neonatal outcomes should account for socioeconomic factors.
Abstract
Importance
Previous studies suggest that race or ethnicity may be associated with risk for developing retinopathy of prematurity (ROP). Little is known about how socioeconomic factors mediate the relationship between race or ethnicity and ROP outcomes.
Objective
To evaluate how socioeconomic factors, in the context of race and ethnicity, are associated with ROP outcomes.
Design, Setting, and Participants
This retrospective cohort study used US Census Bureau income data and electronic medical records from neonatal intensive care units at 4 hospitals, UCLA Mattel Children’s Hospital, UCLA Santa Monica Hospital, Cedars-Sinai Medical Center, and Harbor-UCLA Medical Center. Eligible participants included neonates born at a gestational age (GA) of 30 weeks or less, birth weight less than 1500 g, or a GA at birth greater than 30 weeks but with an unstable clinical course. Participants were screened for ROP between January 1, 2010, and December 31, 2020.
Exposures
Race and ethnicity data, GA, demographic and clinical information, proxy household income, and health insurance status were collected as risk factors.
Main Outcomes and Measures
Diagnosis and severity of ROP were the main study outcomes. Severity was determined according to a classification system developed by the Early Treatment for Retinopathy of Prematurity Cooperative Group.
Results
In a crude model, Hispanic neonates were more likely to be diagnosed with ROP (OR, 1.70; 95% CI, 1.20-2.42) and had more severe ROP (OR, 2.24; 95% CI, 1.21-4.15) compared with non-Hispanic White neonates; these associations were no longer found when adjusting for GA and socioeconomic factors (OR, 1.12; 95% CI, 0.68-1.82, and OR, 1.67; 95% CI, 0.80-3.52, for ROP diagnosis and severity, respectively). In a fully adjusted model, lower GA was the primary predictor of ROP incidence (OR, 0.52; 95% CI, 0.48-0.57; P < .001), and higher median household income was associated with higher GA (OR, 0.26; 95% CI, 0.09-0.43; P = .002).
Conclusions and Relevance
In this cohort study, GA was the primary driver of disparities in ROP outcomes in a heterogeneous population of neonates in Los Angeles, California. When examined in the context of socioeconomic factors, GA did not differ between racial and ethnic groups. Studies of disparities associated with race and ethnicity should consider these constructs in conjunction with other sociodemographic factors and social determinants of health.
Introduction
Retinopathy of prematurity (ROP) is a leading cause of permanent visual impairment in children.1,2,3 While early gestational age (GA) (<30 weeks) and low birth weight (<1500 g) are consistently identified as the strongest risk factors for developing ROP, previous studies have also suggested that race and ethnicity may also be associated with risk for developing ROP.1,4 Past studies have reported that Hispanic neonates have higher rates of severe ROP compared with non-Hispanic neonates, and non-Hispanic Black neonates have lower rates of ROP than non-Hispanic White neonates.4,5,6,7,8 However, these studies did not investigate how socioeconomic factors contribute to the racial or ethnic inequities seen in ROP outcomes.
There is growing awareness that racial and ethnic differences in health outcomes may be due to socioeconomic determinants of health. As such, race and ethnicity must be considered as social factors, not solely biologic.9 Race and ethnicity reflect generations of social and health differences that have contributed to increased risk of disease and poorer health outcomes. As such, there remains a gap in our understanding of how socioeconomic factors are associated with ROP. Our study explores how socioeconomic factors caused by structural inequalities are associated with ROP incidence and severity. We hypothesize that underserved populations have worse ROP outcomes due to modifiable risk factors. In this study, we aim to determine the association between factors such as median household income, health insurance status, and race and ethnicity as they relate to risk of ROP diagnosis and ROP severity.
Methods
This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline. The University of California, Los Angeles (UCLA), institutional review board approved the study protocol and granted waiver of consent.
Study Participants
A multicenter retrospective cohort study was performed at UCLA Mattel Children’s Hospital, UCLA Santa Monica Hospital, Cedars-Sinai Medical Center, and Harbor-UCLA Medical Center with 1267 infants screened for ROP in the neonatal intensive care units (NICUs). All neonates screened for ROP while hospitalized at these institutions were eligible. Study inclusion criteria were consistent with the American Academy of Pediatrics guidelines for ROP screening: infants born at a GA of 30 weeks or less, birth weight less than 1500 g, or GA at birth greater than 30 weeks but with an unstable clinical course deemed high risk for ROP by the medical team.10 Participants were eligible for the study if their record contained all the following demographic and socioeconomic data: race and ethnicity, health insurance status, and maternal zip code. Participants who had an ROP examination but no demographic or socioeconomic data were excluded (eFigure in the Supplement). Participants who were screened for ROP but died before final ROP diagnosis was determined were excluded. Data for 495 neonates were collected from UCLA Mattel Children’s Hospital and UCLA Santa Monica Hospital, academic medical centers, between January 1, 2010, and December 31, 2020. Data for 593 neonates were collected from Cedars-Sinai Medical Center, a community hospital, between January 1, 2010, and December 31, 2020. Data for 146 neonates were collected from Harbor-UCLA Medical Center, a Los Angeles County hospital, between January 1, 2015, and December 31, 2020. Because of limited access to paper records, data from Harbor-UCLA Medical Center between January 1, 2010, and December 31, 2014, were not included.
Demographic and Clinical Data
Demographic (sex, race, ethnicity, primary language, zip code, health insurance) and clinical data (GA, birth weight, bronchopulmonary dysplasia, death) were collected for each patient via electronic medical review. Race and ethnicity information were self-reported by patients; the person giving birth reported if they identified as Asian, Black, White, or unknown or other and then reported if they identified as Hispanic or non-Hispanic. Patients who identified as 1 known race plus other race were categorized as their known race and ethnicity (n = 141). Patients who identified as 2 or more races or ethnicities without indicating a primary identification were marked as other (n = 3). Neonates were categorized into 5 groups based on reported race and ethnicity: Hispanic, non-Hispanic Asian, non-Hispanic Black, non-Hispanic White, and non-Hispanic unknown/other race. Other race included patients who self-reported as Pacific Islander and Native American.
Health insurance at the time of admission to the NICU was classified as private or public (Medi-Cal, Medicaid, and My Health LA). Proxy for household income was determined for each neonate using maternal zip code (neighborhood income level). Data from the US Census Bureau were used to record median household income by each neonates’ reported zip code.11,12 Bronchopulmonary dysplasia was defined as the need for supplemental oxygen or respiratory support at 36 weeks’ corrected GA.13
Board-certified ophthalmologists evaluated study participants for ROP at the recommended intervals according to the 2013 American Academy of Pediatrics guidelines.10 Data were collected via electronic medical review about ROP stage, ROP zone, presence of plus disease, and need for interventional treatment (anti–vascular endothelial growth factor [VEGF] intravitreal injections and/or peripheral retinal ablation). Severity of ROP was determined using the participants’ worst ROP examination with a classification system developed by the Early Treatment for Retinopathy of Prematurity Cooperative Group.14 Under these criteria, participants were classified as having no ROP, low-grade ROP, or type 1 ROP. Low-grade ROP refers to any diagnosis less severe than type 1 ROP, including zone 1 with stage 1 or 2 without plus disease and zone 2 or 3 with stage 1, 2, or 3 without plus disease. Neonates were treated for ROP with peripheral retinal ablation, anti-VEGF intravitreal injections, or both for type 1 ROP per the discretion of the treating ophthalmologist. If infants had persistent low-grade ROP after anti-VEGF injection beyond 52 weeks’ corrected GA, peripheral laser was considered to minimize the burden of follow-up.
Statistical Analyses
Continuous variables are summarized using median and IQRs, and categorical variables are summarized using counts and percentages. Kruskal-Wallis and Fisher exact tests were used to evaluate baseline differences in continuous and categorical variables between neonates diagnosed with no ROP, low-grade ROP, or type 1 ROP. Infants with a reported race of unknown or other were grouped together because of small sample sizes. Median income was estimated based on median household income for each reported zip code and was scaled to its SD because of the wide range of income. Unadjusted and multivariable logistic regressions were used to model dichotomized variables: ROP (yes vs no) in the entire cohort and type 1 ROP vs low-grade in the subpopulation of ROP infants. The association between race and ethnicity and outcome variables was assessed in 3 different models: (1) a model including only race and ethnicity, (2) a model including race and ethnicity and GA in weeks as a continuous variable, and (3) a final model including race and ethnicity, GA, insurance status, and scaled median income.
We also investigated the association between GA and race and ethnicity in this cohort. Gestational age was modeled as the outcome variable of linear regression models, first using race and ethnicity alone, next including insurance status and scaled median income. These steps were repeated for both the entire cohort and the subset of infants with ROP. When modeling ROP (yes vs no) and GA in the entire cohort, birth hospital was modeled as a random effect to account for the multicenter nature of this data set. In the subset analyses of ROP infants, models did not include random effects because of smaller sample size. All tests were 2-sided. Analyses were performed using R version 4.1.0.
Results
In this retrospective cohort study performed at UCLA Mattel Children’s Hospital, UCLA Santa Monica Hospital, Cedars-Sinai Medical Center, and Harbor-UCLA Medical Center, 1267 infants were screened for ROP in the NICU between January 1, 2010, and December 31, 2020. Of the 1267 infants screened for ROP, 1234 neonates had available demographic and socioeconomic data and met inclusion criteria.
Demographic and Clinical Data
Our study cohort of 1234 infants had a median (IQR) GA of 29 weeks (26.9-30.7 weeks) and median (IQR) birth weight of 1150 g (870-1390 g). Five hundred ninety-seven infants (48%) were female, and 388 infants (31%) had bronchopulmonary dysplasia. Our cohort had a median proxy household income of $75 714 (IQR, $51 850-$95 696). Seven hundred sixty-one neonates (62%) had private health insurance, and 473 (38%) had public health insurance. Three hundred forty infants (28%) were identified as Hispanic, 117 non-Hispanic Asian (9%), 219 non-Hispanic Black (18%), 463 non-Hispanic White (38%), and 95 non-Hispanic unknown or other race (8%). One hundred infants (8%) developed type 1 ROP, 206 infants (17%) developed low-grade ROP, and 928 infants (75%) had no ROP. One hundred thirteen infants (9%) were treated for ROP; this included infants with type 1 ROP and those with persistent low-grade ROP. Demographic and clinical information are represented in Table 1.
Table 1. Demographic Information by ROP Type.
| No. (%) | P valuea | ||||
|---|---|---|---|---|---|
| Total cohort (N = 1234) | No ROP (n = 928) | Low-grade ROP (n = 206) | Type 1 ROP (n = 100) | ||
| Gestational age, median (IQR), wk | 29 (26.86-30.71) | 29.7 (28.1-31.0) | 26.6 (25.0-28.5) | 25.3 (24.1-26.0) | <.001 |
| Birth weight, median (IQR), g | 1150 (870-1390) | 1245 (1021-1435) | 833 (671-1087) | 680 (585-780) | <.001 |
| Sex | .10 | ||||
| Male | 637 (52) | 495 (53.3) | 97 (47.1) | 45 (45.0) | |
| Female | 597 (48) | 433 (46.7) | 109 (52.9) | 55 (55.0) | |
| Diagnosis of BPD | 388 (31) | 187 (20.2) | 125 (60.7) | 76 (76.0) | <.001 |
| Household income, median (IQR), thousands of dollars/yb | 75.71 (51.85-95.7) | 78.5 (54.7-96.6) | 64.6 (49.1-97.3) | 55.1 (45.9-77.8) | <.001 |
| Health insurance | <.001 | ||||
| Public | 473 (38) | 307 (33.1) | 105 (50.9) | 61 (61.0) | |
| Private | 761 (62) | 621 (66.9) | 101 (49.0) | 39 (39.0) | |
| Race and ethnicity | <.001 | ||||
| Hispanic | 340 (28) | 227 (24.5) | 66 (32.0) | 47 (47.0) | |
| Non-Hispanic Asian | 117 (9) | 96 (10.3) | 13 (6.3) | 8 (8.0) | |
| Non-Hispanic Black | 219 (18) | 165 (17.8) | 41 (19.9) | 13 (13.0) | |
| Non-Hispanic White | 463 (38) | 376 (40.5) | 66 (32.0) | 21 (21.0) | |
| Unknown/other | 95 (8) | 64 (6.9) | 20 (9.7) | 11 (11.0) | |
Abbreviations: BPD, bronchopulmonary dysplasia; ROP, retinopathy of prematurity.
A Fisher exact test was used to calculate P values comparing categorical variables summarized by No. (%), and a Kruskal-Wallis test was used to compare continuous variables summarized by median and IQRs across levels of ROP diagnosis.
Data provided by the US Census Bureau was used for median household income by each reported zip code.
Gestational age varied by race and ethnicity (P = .001) (eTable 1 in the Supplement). Hispanic neonates were born a mean 0.66 weeks earlier than non-Hispanic White neonates (point estimate [PE] −0.66; 95% CI, −1.01 to −0.30), and Black neonates were born 0.67 weeks earlier than non-Hispanic White neonates (PE, −0.67; 95% CI, −1.07 to −0.26) (eTable 2 in the Supplement). Similarly, birth weight varied by race and ethnicity (P = .001) (eTable 1 in the Supplement). Specifically, Hispanic neonates had birth weights that were a mean 56.45 g less than non-Hispanic white neonates (PE, −56.45; 95% CI, −107.11 to −5.79) and Black neonates had birth weights that were a mean 118.76 g less than non-Hispanic white neonates (PE, −118.76; 95% CI, −176.93 to −60.58) (eTable 2 in the Supplement). Health insurance type also varied by race and ethnicity (P < .001) (eTable 1 in the Supplement). Hispanic neonates had 12.08 times the odds of having public health insurance when compared with non-Hispanic white neonates (odds ratio [OR], 12.08; 95% CI, 8.55-17.29). Similarly, Black neonates had 11.44 times the odds of having public health insurance when compared with non-Hispanic White neonates (OR, 11.44; 95% CI, 7.82-16.95) (eTable 2 in the Supplement). The mean proxy household income for non-Hispanic White neonates was $29 190 greater than Hispanic neonates’ household income (PE, −29.19; 95% CI, −32.98 to −25.39) and $31 140 greater than non-Hispanic Black neonates’ household income (PE, −31.14; 95% CI, −35.49 to −26.78) (eTable 2 in the Supplement).
ROP Diagnosis
In the unadjusted model, race and ethnicity were associated with a diagnosis of ROP (including low-grade and type 1 ROP) (P = .03). Hispanic neonates had 1.70 times the odds of having a diagnosis of ROP when compared with non-Hispanic White neonates (OR, 1.70; 95% CI, 1.20-2.42), and Black neonates had 1.49 times the odds of having a diagnosis of ROP when compared with non-Hispanic White neonates (OR, 1.49; 95% CI, 0.98-2.27) (Table 2, model 1). After adding GA to the model, race and ethnicity were no longer associated with diagnosis of ROP (P = .40); specifically Hispanic neonates were not more likely to be diagnosed with ROP compared with non-Hispanic White neonates (OR, 1.28; 95% CI, 0.83-1.97), and only GA remained associated with diagnosis of ROP (OR, 0.52; 95% CI, 0.48-0.56; P < .001) (Table 2, model 2). Next, in the fully adjusted model, after adding markers of socioeconomic status, income and health insurance status, along with GA and race and ethnicity, insurance status (OR, 1.39; 95% CI, 0.88-2.21; P = .16) and median household income (OR, 1.00; 95% CI, 0.81-1.23; P = .99) were not associated with increased risk of ROP while GA remained associated with the diagnosis of ROP (OR, 0.52; 95% CI, 0.48-0.57; P < .001) (Table 2, model 3).
Table 2. Modeling ROP Diagnosis as Outcome Variable Using Unadjusted and Adjusted Logistic Regression.
| Odds ratio (95% CI) | P valuea | |
|---|---|---|
| Model 1b | ||
| Race and ethnicity | .03 | |
| Non-Hispanic White | 1 [Reference] | |
| Hispanic | 1.70 (1.20-2.42) | |
| Non-Hispanic Asian | 0.95 (0.55-1.45) | |
| Non-Hispanic Black | 1.49 (0.98-2.27) | |
| Unknown/other | 1.31 (0.50-2.20) | |
| Model 2c | ||
| Gestational age (1-wk increase) | 0.52 (0.48-0.56) | <.001 |
| Race and ethnicity | .40 | |
| Non-Hispanic White | 1 [Reference] | |
| Hispanic | 1.28 (0.83-1.97) | |
| Non-Hispanic Asian | 0.84 (0.43-1.64) | |
| Non-Hispanic Black | 0.92 (0.55-1.54) | |
| Unknown/other | 1.52 (0.80-2.88) | |
| Model 3d | ||
| Gestational age (1-wk increase) | 0.52 (0.48-0.57) | <.001 |
| Health insurance | .16 | |
| Private | 1 [Reference] | |
| Public | 1.39 (0.88-2.21) | |
| Income (1-SD increase) | 1.00 (0.81-1.23) | .99 |
| Race and ethnicity | .54 | |
| Non-Hispanic White | 1 [Reference] | |
| Hispanic | 1.12 (0.68-1.82) | |
| Non-Hispanic Asian | 0.86 (0.44-1.67) | |
| Non-Hispanic Black | 0.81 (0.46-1.43) | |
| Unknown/other | 1.41 (0.74-2.71) | |
Abbreviation: ROP, retinopathy of prematurity.
Calculated using a Wald test.
Model 1: unadjusted model with ROP diagnosis as outcome variable of race and ethnicity.
Model 2: adjusted model with ROP diagnosis as outcome variable of race and ethnicity and gestational age.
Model 3: adjusted model with ROP diagnosis as outcome variable of race and ethnicity, gestational age, insurance status, and median household income.
Gestational Age and ROP Diagnosis
Earlier GAs at birth are more closely associated with socioeconomic factors than race and ethnicity in this cohort. In the unadjusted model, race and ethnicity were associated with an earlier GA (P = .001), with Hispanic (OR, −0.66; 95% CI, −1.01 to −0.30) and non-Hispanic Black neonates (OR, −0.67; 95% CI, −1.07 to −0.26) born earlier than non-Hispanic White neonates (Table 3, model 1). After adjusting for median household income and insurance status as a proxy for socioeconomic status, race and ethnicity were no longer a significant predictor of GA (P = .40). However, lower household income was associated with lower GA (OR, 0.26; 95% CI, 0.09 to 0.43; P = .002), though insurance type was not associated with GA (OR, −0.26; 95% CI, −0.61 to 0.09; P = .14) (Table 3, model 2).
Table 3. Modeling Gestational Age as the Outcome Variable Using Unadjusted and Adjusted Linear Regression in the Full Cohort.
| Point estimate (95% CI) | P valuea | |
|---|---|---|
| Model 1b | ||
| Race and ethnicity | .001 | |
| Non-Hispanic White | 1 [Reference] | |
| Hispanic | −0.66 (−1.01 to −0.30) | |
| Non-Hispanic Asian | −0.06 (−0.58 to 0.35) | |
| Non-Hispanic Black | −0.67 (−1.07 to −0.26) | |
| Unknown/other | −0.02 (−0.62 to 0.54) | |
| Model 2c | ||
| Health insurance | .14 | |
| Private | 1 [Reference] | |
| Public | −0.26 (−0.61 to 0.09) | |
| Income (1-SD increase) | 0.26 (0.09 to 0.43) | .002 |
| Race and ethnicity | .40 | |
| Non-Hispanic White | 1 [Reference] | |
| Hispanic | −0.27 (−0.68 to 0.13) | |
| Non-Hispanic Asian | 0.01 (−0.5 to 0.52) | |
| Non-Hispanic Black | −0.27 (−0.72 to 0.18) | |
| Unknown/other | 0.22 (−0.35 to 0.79) | |
Calculated using a Wald test.
Model 1: unadjusted model with gestational age (weeks) as outcome variable of race and ethnicity.
Model 2: adjusted model with gestational age (weeks) as outcome variable of race and ethnicity, insurance status, and household income.
ROP Severity
In the unadjusted model for ROP severity comparing low-grade ROP to type 1 ROP, race and ethnicity were not a predictor of ROP severity. However, Hispanic neonates had 2.24 times the odds to have more severe ROP (type 1 ROP) when compared with non-Hispanic White neonates (OR, 2.24; 95% CI, 1.21-4.15) (Table 4, model 1). When GA and race and ethnicity were included in a multivariable analysis, Hispanic ethnicity remained associated with an increased risk of developing more severe ROP when compared with non-Hispanic White neonates (OR, 2.14; 95% CI, 1.10-4.13) (Table 4, model 2).
Table 4. Modeling ROP Severity as Outcome Variable Using Unadjusted and Adjusted Logistic Regression.
| Odds ratio (95% CI) | P valuea | |
|---|---|---|
| Model 1b | ||
| Race and ethnicity | .06 | |
| Non-Hispanic White | 1 [Reference] | |
| Hispanic | 2.24 (1.21-4.15) | |
| Non-Hispanic Asian | 1.93 (0.71-4.28) | |
| Non-Hispanic Black | 1.00 (0.45-2.20) | |
| Unknown/other | 1.73 (0.80-4.19) | |
| Model 2c | ||
| Gestational age (1-wk increase) | 0.65 (0.56-0.76) | <.001 |
| Race and ethnicity | .02 | |
| Non-Hispanic White | 1 [Reference] | |
| Hispanic | 2.14 (1.10-4.13) | |
| Non-Hispanic Asian | 1.81 (0.61-5.40) | |
| Non-Hispanic Black | 0.71 (0.31-1.66) | |
| Unknown/other | 2.11 (0.81-5.47) | |
| Model 3d | ||
| Gestational age (1-wk increase) | 0.66 (0.57-0.76) | <.001 |
| Health insurance | .96 | |
| Private | 1 [Reference] | |
| Public | 0.98 (0.53-1.84) | |
| Income (1-SD increase) | 0.75 (0.52-1.08) | .12 |
| Race and ethnicity | .03 | |
| Non-Hispanic White | 1 [Reference] | |
| Hispanic | 1.67 (0.80-3.52) | |
| Non-Hispanic Asian | 1.79 (0.60-5.34) | |
| Non-Hispanic Black | 0.54 (0.21-1.37) | |
| Unknown/other | 1.89 (0.70-5.06) | |
Abbreviation: ROP, retinopathy of prematurity.
Calculated using a Wald test.
Model 1: unadjusted model with ROP severity as outcome variable of race and ethnicity.
Model 2: adjusted model with ROP severity as outcome variable of race and ethnicity and gestational age.
Model 3: adjusted model with ROP severity as outcome variable of race and ethnicity, gestational age, insurance status, and median household income.
When insurance status and median household income were added to GA and race and ethnicity in the model, GA (OR, 0.66; 95% CI, 0.57-0.76; P < .001) remained associated with more severe ROP (Table 4, model 3). However, Hispanic ethnicity was no longer associated with more severe ROP compared with non-Hispanic White neonates (OR, 1.67; 95% CI, 0.80-3.52). When compared with non-Hispanic White neonates, no other race or ethnicity category was associated with worse ROP. Insurance status (OR, 0.98; 95% CI, 0.53-1.84; P = .96) and median household income (OR, 0.75; 95% CI, 0.52-1.08; P = .12) were not associated with more severe ROP (Table 4, model 3).
Gestational Age and ROP Severity
Given that GA is an independent predictor of ROP severity, we again aimed to determine the association between race and ethnicity, insurance status and median household income, and GA within this subset of infants who developed ROP. In the unadjusted model, race was not associated with an earlier GA at birth in this cohort of neonates with ROP (P = .08). However, Black infants were born a mean 0.93 weeks earlier than non-Hispanic White neonates (PE, −0.93; 95% CI, −1.69 to −0.18) (Table 5, model 1). After adjustment for median household income and insurance status as a proxy for socioeconomic status, race and ethnicity were not associated with GA (P = .26), and Black neonates were born a mean 0.53 weeks earlier than non-Hispanic White neonates (PE, −0.53; 95% CI, −1.36 to 0.29) (Table 5, model 2).
Table 5. Modeling Gestational Age as Outcome Variable Using Unadjusted and Adjusted Linear Regression in the Subset of the Cohort Diagnosed With ROP.
| Point estimate (95% CI)a | P valuea | |
|---|---|---|
| Model 1b | ||
| Race and ethnicity | .08 | |
| Non-Hispanic White | 1 [Reference] | |
| Hispanic | −0.42 (−1.04 to 0.20) | |
| Non-Hispanic Asian | −0.31 (−1.38 to 0.45) | |
| Non-Hispanic Black | −0.93 (−1.69 to −0.18) | |
| Unknown/other | 0.28 (−1.23 to 1.19) | |
| Model 2c | ||
| Health insurance | .26 | |
| Private | 1 [Reference] | |
| Public | −0.34 (−0.93 to 0.25) | |
| Income (1-SD increase) | 0.26 (−0.07 to 0.59) | .12 |
| Race and ethnicity | .34 | |
| Non-Hispanic White | 1 [Reference] | |
| Hispanic | −0.03 (−0.73 to 0.67) | |
| Non-Hispanic Asian | −0.29 (−1.35 to 0.77) | |
| Non-Hispanic Black | −0.53 (−1.36 to 0.29) | |
| Unknown/other | 0.49 (−0.44 to 1.42) | |
Abbreviation: ROP, retinopathy of prematurity.
P values and 95% CI were calculated using Wald test.
Model 1: unadjusted model with gestational age (weeks) as outcome variable of race and ethnicity.
Model 2: adjusted model with gestational age (weeks) as outcome variable of race and ethnicity, insurance status, and household income.
Discussion
The major objective of this study was to better understand what socioeconomic risk factors put neonates at risk for ROP, by looking beyond categories of race and ethnicity and to better understand the effect of potential health inequities. We found that Hispanic ethnicity was associated with more severe ROP through lower GA and that Black race was associated with preterm birth through lower household income.
Past studies have found that Hispanic neonates may be 33% more likely to develop ROP.15 Consistent with previous reports, we found that Hispanic neonates were more likely to be diagnosed with ROP than non-Hispanic White neonates.4 However, when we looked at the entire screened population in our study, after adjusting for earlier GA, Hispanic neonates were no longer more likely to be diagnosed with ROP. This indicates that Hispanic ethnicity through lower GA is a risk factor for ROP, and this observation should be further explored with other social determinants of health. It is important to note that birth weight and diagnosis of bronchopulmonary dysplasia were also associated with ROP type (Table 1). To explore this association, we ran a model including birth weight and bronchopulmonary dysplasia in addition to GA, insurance type, income, and race and ethnicity. In this model, the coefficient of the variables did not change from our fully adjusted model (eTable 3 in the Supplement).
When examining factors associated with lower GA in infants screened for ROP, we found that Black neonates were more likely to be born at earlier GAs than non-Hispanic White neonates. However, after we adjusted for household income and insurance status, this association between race and ethnicity and GA was no longer significant. In addition, household income was found to be associated with earlier GA. Therefore, our study emphasizes the role of socioeconomic factors beyond race and ethnicity in determining prematurity and ROP risk. It is important to note that there may be additional clinical risk factors associated with GA, including but not limited to maternal factors such as gestational diabetes, hypertensive disorders, and age. Future studies evaluating these additional clinical risk factors would augment the understanding of socioeconomic determinants of prematurity and ROP.
This study implies that socioeconomic determinants of health, such as insurance status and household income, affect important long-term health outcomes of maternal-neonatal dyads. Our findings are consistent with previous work showing that neighborhood socioeconomic status is an important factor that clinicians should consider to identify mothers at risk for preterm birth.16 Thus, when assessing ROP risk, it is important to understand unseen socioeconomic factors that may be affecting maternal health and preterm birth and subsequently neonatal health to optimize health outcomes.
Limitations
The major limitation of our study was its retrospective nature, limiting which socioeconomic factors we were able to assess. There are inherent challenges in the way race and ethnicity are categorized in the medical record; a few patients in our study identified as other, which makes this group uninterpretable within the confines of our study. In addition, household income was also not self-reported but estimated from the zip code provided by the mother on admission to the hospital. Another limitation of our study was missing data. Some individual records were missing race, ethnicity, income, and insurance data, thus making them ineligible to be included in our study. However, less than 3% of the neonates screened for ROP were excluded for this reason in our study. Importantly, our findings do not prove cause and effect and could be due, at least in part, to other confounding factors. Despite these limitations, we were able to gather and analyze data on a large multicenter cohort of patients from a socioeconomically diverse population in Los Angeles.
Conclusions
Our study evaluating ROP incidence and severity demonstrated the importance of a multifactorial approach to evaluating ROP risk. Future studies from other geographical areas with different health insurance practices would be beneficial to grow our understanding of how structural inequalities are associated with maternal-fetal health outcomes. This study emphasizes the potential importance of intervening on maternal economic stability and prenatal access to care, which are modifiable risk factors.
eFigure. Flow chart representing study cohort missingness
eTable 1. Cohort characteristics stratified by race and ethnicity
eTable 2. Comparisons of cohort characteristics by race and ethnicity
eTable 3. Modeling ROP diagnosis as outcome variable using unadjusted and adjusted logistic regressions including BPD and birth weight
References
- 1.Bashinsky AL. Retinopathy of prematurity. N C Med J. 2017;78(2):124-128. doi: 10.18043/ncm.78.2.124 [DOI] [PubMed] [Google Scholar]
- 2.Hirvonen M, Ojala R, Korhonen P, et al. Visual and hearing impairments after preterm birth. Pediatrics. 2018;142(2):e20173888. doi: 10.1542/peds.2017-3888 [DOI] [PubMed] [Google Scholar]
- 3.Kim SJ, Port AD, Swan R, Campbell JP, Chan RVP, Chiang MF. Retinopathy of prematurity: a review of risk factors and their clinical significance. Surv Ophthalmol. 2018;63(5):618-637. doi: 10.1016/j.survophthal.2018.04.002 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Ying GS, Bell EF, Donohue P, Tomlinson LA, Binenbaum G; G-ROP Research Group . Perinatal risk factors for the retinopathy of prematurity in postnatal growth and ROP study. Ophthalmic Epidemiol. 2019;26(4):270-278. doi: 10.1080/09286586.2019.1606259 [DOI] [PubMed] [Google Scholar]
- 5.Husain SM, Sinha AK, Bunce C, et al. Relationships between maternal ethnicity, gestational age, birth weight, weight gain, and severe retinopathy of prematurity [published correction appears in J Pediatr. 2013;163(6):1798]. J Pediatr. 2013;163(1):67-72. doi: 10.1016/j.jpeds.2012.12.038 [DOI] [PubMed] [Google Scholar]
- 6.Gantz MG, Carlo WA, Finer NN, et al. ; SUPPORT Study Group of the Eunice Kennedy Shriver National Institute of Child Health and Human Development Neonatal Research Network . Achieved oxygen saturations and retinopathy of prematurity in extreme preterms. Arch Dis Child Fetal Neonatal Ed. 2020;105(2):138-144. doi: 10.1136/archdischild-2018-316464 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Saunders RA, Donahue ML, Christmann LM, et al. ; The Cryotherapy for Retinopathy of Prematurity Cooperative Group . Racial variation in retinopathy of prematurity. Arch Ophthalmol. 1997;115(5):604-608. doi: 10.1001/archopht.1997.01100150606005 [DOI] [PubMed] [Google Scholar]
- 8.McGovern AM, Greenspan JS, Webb D, Kirkby S, Culhane JF, Desai S. Retinopathy of prematurity: does race matter? J Neonatal Perinatal Med. 2009;2(3):157-162. doi: 10.3233/NPM-2009-0063 [DOI] [Google Scholar]
- 9.Amutah C, Greenidge K, Mante A, et al. Misrepresenting race: the role of medical schools in propagating physician bias. N Engl J Med. 2021;384(9):872-878. doi: 10.1056/NEJMms2025768 [DOI] [PubMed] [Google Scholar]
- 10.Fierson WM; American Academy of Pediatrics Section on Ophthalmology; American Academy of Ophthalmology; American Association for Pediatric Ophthalmology and Strabismus; American Association of Certified Orthoptists . Screening examination of premature infants for retinopathy of prematurity. Pediatrics. 2013;131(1):189-195. doi: 10.1542/peds.2012-2996 [DOI] [PubMed] [Google Scholar]
- 11.US Census Bureau . Explore Census data. Accessed July 2021.https://data.census.gov/cedsci/
- 12.Berkowitz SA, Traore CY, Singer DE, Atlas SJ. Evaluating area-based socioeconomic status indicators for monitoring disparities within health care systems: results from a primary care network. Health Serv Res. 2015;50(2):398-417. doi: 10.1111/1475-6773.12229 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Jensen EA, Dysart K, Gantz MG, et al. The diagnosis of bronchopulmonary dysplasia in very preterm infants: an evidence-based approach. Am J Respir Crit Care Med. 2019;200(6):751-759. doi: 10.1164/rccm.201812-2348OC [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Early Treatment for Retinopathy of Prematurity Cooperative Group . Revised indications for the treatment of retinopathy of prematurity: results of the Early Treatment for Retinopathy of Prematurity randomized trial. Arch Ophthalmol. 2003;121(12):1684-1694. doi: 10.1001/archopht.121.12.1684 [DOI] [PubMed] [Google Scholar]
- 15.Lad EM, Nguyen TC, Morton JM, Moshfeghi DM. Retinopathy of prematurity in the United States [published correction appears in Br J Ophthalmol. 2010;94(9):1268]. Br J Ophthalmol. 2008;92(3):320-325. doi: 10.1136/bjo.2007.126201 [DOI] [PubMed] [Google Scholar]
- 16.Boubred F, Pauly V, Romain F, Fond G, Boyer L. The role of neighbourhood socioeconomic status in large for gestational age. PLoS One. 2020;15(6):e0233416. doi: 10.1371/journal.pone.0233416 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eFigure. Flow chart representing study cohort missingness
eTable 1. Cohort characteristics stratified by race and ethnicity
eTable 2. Comparisons of cohort characteristics by race and ethnicity
eTable 3. Modeling ROP diagnosis as outcome variable using unadjusted and adjusted logistic regressions including BPD and birth weight