Delays and loss to follow-up in outpatient retinopathy of prematurity screening

Abstract

PURPOSE

To determine the frequency and determinants of delayed and complete loss to follow-up (LTFU) for the first outpatient retinopathy of prematurity (ROP) examination after discharge at a single tertiary center.

METHODS

A retrospective chart review of neonates at a single institution from 2014 to 2023 who underwent inpatient ROP screening and were intended to have outpatient follow-up there within 1 month was performed. Frequency of LTFU and delayed follow-up (>3 days) were calculated, and univariate logistic regression revealed potential associations with clinical and sociodemographic factors. Qualitative themes were extracted to identify reasons for gaps.

RESULTS

Of the 721 infants who met inclusion criteria, 691 (95.8%) received follow-up, although 12.9% of those receiving follow-up were delayed by a median of 9 days. Zone III ROP or retinal maturity on inpatient screening was associated with complete LTFU (OR = 0.20; 95% CI, 0.09 – 0.45). Postmenstrual age at discharge ≥40 weeks was also associated with LTFU (OR = 0.31; 95% CI, 0.14-0.65), but among those with follow-up, it was more likely to be timely (OR = 1.84; 95% CI, 1.08-3.30). Inpatient ROP treatment was associated with timeliness of follow-up (OR = 5.27; 95% CI, 1.11-94.2); Hispanic or non-White race/ethnicity was associated with delays (OR = 0.44; 95% CI, 0.28-0.69).

CONCLUSIONS

Despite an overall high rate of follow-up, delays remain a challenge that may disproportionately affect Hispanic or non-White patients. ROP severity may be associated with better follow-up. Although these findings are exploratory, they may inform future research and interventions to ensure uninterrupted care for all patients. (J AAPOS 2026;30: 104793)

Retinopathy of prematurity (ROP) is a leading cause of preventable childhood blindness and vision impairment globally.¹ Development of retinal vasculature is incomplete in premature infants, and subsequent exposure to high supplemental oxygen typically further compromises the immature vessels.¹ In the United States, ROP incidence rates among premature births appear to have increased by 86% from 2003 to 2019, with the largest increase observed in minority groups.² Current US guidelines recommend screening for neonates at risk by birth weight, gestational age, or tenuous medical course, with repeated examinations until the retina is fully vascularized or risk subsides.³

Without timely screening and management, ROP may progress to severe vision loss or blindness.⁴ Consistent follow-up is critical for preventing adverse outcomes. This urgency, combined with the complex logistics of screening premature infants and the frequent need for multiple examinations, creates numerous opportunities for care discontinuity.⁵

Despite advances in screening protocols and treatment modalities, delays and loss to follow-up (LTFU) continue to present a significant barrier to optimal outcomes.^6,7 Understanding and addressing the factors contributing to LTFU in ROP care represents a critical opportunity to preserve vision in this vulnerable population. A study at a tertiary medical center showed that maternal ethnicity, family income lower than the national average, lower gestational age, and lower birth weight were associated with lower adherence in univariate analyses, though after adjusting for possible confounders in multivariable analyses, only lower birth weight remained significant.⁸

Although previous studies have identified sociodemographic and clinical associations with absolute adherence, they have largely been small scale studies, and these patient factors may vary greatly across institutions. Though studies report mixed findings, for the majority of patients who receive follow-up, less is known about the appropriate timeliness of their care.^8,9 This study aimed to describe the levels of timely follow-up, delayed follow-up, and complete LTFU, and determinants of follow-up status for the first outpatient ROP examination after hospital discharge at a single tertiary center from 2014 to 2023.

Subjects and methods

This study was determined to be exempt from approval by the Institutional Review Board of the University of Michigan. Study protocols adhered to the tenets of the Declaration of Helsinki and complied with requirements of the US Health Insurance Portability and Accountability Act of 1996. The medical records of 1,397 premature infants admitted to the neonatal intensive care unit (NICU) at the University of Michigan between February 2014 and December 2023 who met screening criteria for ROP according to the American Academy of Pediatrics screening guidelines were reviewed retrospectively. These criteria are identical to institutional guidelines and recommend screening neonates with birth weight of 1,500 g or below, gestational age of ≤30 weeks of age, or those with tenuous medical status.³ Infants who underwent ROP screening at least once while inpatient and were recommended to have outpatient follow-up at the same institution within 1 month were included in the study. Patients who never received an inpatient exam, were transferred to another hospital, had planned follow-up elsewhere, or were deceased prior to outpatient follow-up were excluded (Supplement 1, available at jaapos.org, Figure S1).

Data was collected from the last inpatient ROP examination conducted during the admission. If the infant was discharged and readmitted prior to the recommended follow-up date, the subsequent admission data was used instead. The typical NICU discharge course at the institution is as follows: the ROP coordinator may round with the primary team to reinforce family education on the significance of ROP and follow-up. A follow-up appointment is made based on the physician-recommended interval and family’s preferred location; families are not responsible for scheduling. On discharge, the primary team reminds the family about all follow-up items, including ROP examinations. Though this is not a formal protocol, having one assigned ROP coordinator ensures consistency. For patients who miss outpatient ROP appointments, a patient services associate contacts the parent/guardian multiple times by phone and/or portal message to reschedule. If the caregiver does not respond or is not adherent to follow-up, the ophthalmologist may escalate intervention through a certified letter or social work or Child Protective Services (CPS) involvement, depending on the severity of disease.

Delayed follow-up was defined as a delay in outpatient follow-up of more than 3 days from the recommended date. If there were discrepancies in follow-up recommendations between the examination and discharge note, the date furthest from the inpatient exam was used. LTFU was defined as no appointment within 1 year of the recommended date. Clinical data collected included gestational age, birth weight, and postmenstrual age (PMA) at discharge. Also, ROP stage and zone and presence of plus disease were documented for the worse-seeing eye. Inpatient treatment details were recorded at the patient level. The following sociodemographic data was collected: sex, race, ethnicity, preferred language, primary insurance payer, social deprivation index (SDI), and estimated driving distance from clinic (in minutes).¹⁰ SDI and distance from clinic were both approximated at the zip code level, with higher SDI corresponding to greater disadvantage. For analysis, SDI groups were categorized into top 10% or greatest deprivation and bottom 90% or least deprivation among the patient cohort. Qualitative reasons for delay or LTFU were extracted by thematic analysis of documentation in patient charts, when available.

Statistical analysis was performed with R (R Core Team, R Foundation for Statistical Computing; 2021. https://www.R-project.org/). The frequency of timely, delayed, and loss to follow-up was calculated along with descriptive statistics. Univariate logistic regression was conducted to investigate potential associations with LTFU and timely follow-up, and profile confidence intervals were calculated for the odds ratios. Originally, multivariable regression was planned for variables that showed statistically significant associations in the univariate analyses. However, due to very low numbers for certain categories within relevant variables, this was deemed not to be appropriate for the data. Given low case counts for the outcomes of interest, Poisson regression was also performed as a sensitivity analysis. The Fisher exact test was used to further explore potential associations between follow-up status (timely, delayed, and complete LTFU) and categorical patient characteristics. Qualitative analysis of charts was performed to explore the reasons documented for suboptimal follow-up. P values were compared to the 0.05 significance level.

Results

A total of 721 infants met inclusion criteria (Table 1). Approximately 52.0% (n = 375) of infants were female and 56.9% (n = 410) were non-Hispanic White. Patient families were largely English-speaking (n = 697 [96.7%]), with relatively few covered by Medicaid as their primary form of insurance (n = 107 [14.8%]). (This was also true of the patients whose charts were reviewed but did not meet inclusion criteria: 96.7% (n = 654) of the 676 excluded patients had English as their preferred language, and 16.6% (n = 112) of them had Medicaid as their primary form of insurance.) Most included infants had a gestational age ≤30 weeks (n = 596 [82.7%]) and birth-weight ≤1,500 g (n = 662 [91.8%]) (Table 2). Clinically, most infants had ROP stage 0 at the time of their last inpatient examination (n = 581 [80.6%]), and most of their ROP was in zone II (n = 615 [85.3%]) with only 2 (0.3%) exhibiting plus disease. A total of 36 infants (5.0%) received treatment while inpatient. Of these, 24 received anti-VEGF; 20, laser; and 5, vitrectomy—in some cases with multiple treatments in combination, as shown in Table 1. The mean PMA at discharge was approximately 38.4 weeks.

Table 1.

Participant characteristics (N = 721)

Characteristic	Overalla
Sex
Female	375 (52.0)
Male	346 (48.0)
Race/ethnicity
Hispanic, any race	35 (4.9)
Non-Hispanic Asian	37 (5.1)
Non-Hispanic Black	137 (19.0)
Non-Hispanic Other	76 (10.5)
Non-Hispanic White	410 (56.9)
Missing	26 (3.6)
Preferred language
English	697 (96.7)
Non-English	24 (3.3)
Insurance
Medicaid	107 (14.8)
Other	614 (85.2)
Social Deprivation Index score
Mean ± SD	41.8 ± 28.7
Median [min, max]	40.0 [1.00, 99.0]
Missing	8 (1.1)
Driving distance (minutes)
Mean ± SD	48.6 ± 55.1
Median [min, max]	35.4 [8.30, 523]
Missing	32 (4.4)
Stage at last inpatient exam
0	581 (80.6)
1	14 (1.9)
2	102 (14.1)
3	6 (0.8)
4a	2 (0.3)
4b	0 (0)
5	1 (0.1)
Regressed	13 (1.8)
Mature	1 (0.1)
Missing	1 (0.1)
Zone at last inpatient exam
I	0 (0)
II	615 (85.3)
III	96 (13.3)
Mature	1 (0.1)
Missing	9 (1.2)
Plus disease at last inpatient exam
Absent	713 (98.9)
Present	2 (0.3)
Missing	6 (0.8)
Inpatient ROP treatment
No	685 (95.0)
Yes	36 (5.0)
Treatment type
Anti-VEGF	15 (2.1)
Anti-VEGF and laser	7 (1.0)
Anti-VEGF and vitrectomy	1 (0.1)
Anti-VEGF, laser, and vitrectomy	1 (0.1)
Laser	9 (1.2)
Laser and vitrectomy	3 (0.4)
No treatment	685 (95.0)
GA (completed weeks)
Mean ± SD	28.3 ± 2.49
Median [min, max]	29.0 [22.0, 37.0]
PMA at time of discharge (completed weeks)
Mean ± SD	38.4 ± 3.53
Median [min, max]	38.0 [33.0, 60.0]
Birth weight, g
Mean ± SD	1130 ± 312
Median [min, max]	1140 [68.0, 2840]
Missing	1 (0.1)

GA, gestational age; PMA, postmenstrual age; ROP retinopathy of prematurity.

^aNumber (%) unless otherwise stated.

Table 2.

Follow-up status stratified by participant characteristics^a

Characteristic	Follow-up			Overall (n = 721)	P valueb
	Timely (n = 602)	Delayed (n = 89)	Lost to (n = 30)
Sex					0.77
Female	308 (51.2)	51 (57.3)	16 (53.3)	375 (52.0)
Male	294 (48.8)	38 (42.7)	14 (46.7)	346 (48.0)
Race/ethnicity					0.017
Hispanic, any race	30 (5.0)	3 (3.4)	2 (6.7)	35 (4.9)
Non-Hispanic Asian	30 (5.0)	5 (5.6)	2 (6.7)	37 (5.1)
Non-Hispanic Black	100 (16.6)	27 (30.3)	10 (33.3)	137 (19.0)
Non-Hispanic Other	60 (10.0)	15 (16.9)	1 (3.3)	76 (10.5)
Non-Hispanic White	361 (60.0)	36 (40.4)	13 (43.3)	410 (56.9)
Missing	21 (3.5)	3 (3.4)	2 (6.7)	26 (3.6)
Preferred language					0.93
English	581 (96.5)	86 (96.6)	30 (100)	697 (96.7)
Non-English	21 (3.5)	3 (3.4)	0 (0)	24 (3.3)
Insurance					0.38
Medicaid	84 (14.0)	16 (18.0)	7 (23.3)	107 (14.8)
Other	518 (86.0)	73 (82.0)	23 (76.7)	614 (85.2)
Social Deprivation Index					0.62
Less deprivation (bottom 90)	537 (89.2)	80 (89.9)	25 (83.3)	642 (89.0)
High deprivation (top 10)	57 (9.5)	9 (10.1)	5 (16.7)	71 (9.8)
Missing	8 (1.3)	0 (0)	0 (0)	8 (1.1)
Stage of ROP at last inpatient exam					0.71
0	481 (79.9)	71 (79.8)	29 (96.7)	581 (80.6)
1	12 (2.0)	2 (2.2)	0 (0)	14 (1.9)
2	89 (14.8)	13 (14.6)	0 (0)	102 (14.1)
3	5 (0.8)	1 (1.1)	0 (0)	6 (0.8)
4a	2 (0.3)	0 (0)	0 (0)	2 (0.3)
4b	0 (0)	0 (0)	0 (0)	0 (0)
5	1 (0.2)	0 (0)	0 (0)	1 (0.1)
Regressed	12 (2.0)	1 (1.1)	0 (0)	13 (1.8)
Mature	0 (0)	1 (1.1)	0 (0)	1 (0.1)
Missing	0 (0)	0 (0)	1 (3.3)	1 (0.1)
Zone of ROP at last inpatient exam					0.002
I	0 (0)	0 (0)	0 (0)	0 (0)
II	523 (86.9)	75 (84.3)	17 (56.7)	615 (85.3)
III	72 (12.0)	12 (13.5)	12 (40.0)	96 (13.3)
Mature	0 (0)	1 (1.1)	0 (0)	1 (0.1)
Missing	7 (1.2)	1 (1.1)	1 (3.3)	9 (1.2)
Plus disease at last inpatient exam					.0.99
Absent	595 (98.8)	88 (98.9)	30 (100)	713 (98.9)
Present	2 (0.3)	0 (0)	0 (0)	2 (0.3)
Missing	5 (0.8)	1 (1.1)	0 (0)	6 (0.8)
Inpatient ROP treatment					0.29
No	568 (94.4)	88 (98.9)	29 (96.7)	685 (95.0)
Yes	34 (5.6)	1 (1.1)	1 (3.3)	36 (5.0)
GA					0.44
≤30 completed weeks or less	497 (82.6)	77 (86.5)	22 (73.3)	596 (82.7)
>30 wks	105 (17.4)	12 (13.5)	8 (26.7)	125 (17.3)
PMA at discharge (completed wks)					0.002
<40 completed wks	420 (69.8)	72 (80.9)	13 (43.3)	505 (70.0)
≥40 completed	182 (30.2)	17 (19.1)	17 (56.7)	216 (30.0)
BW					0.88
>1500 g	47 (7.8)	9 (10.1)	2 (6.7)	58 (8.0)
≤1500 g	554 (92.0)	80 (89.9)	28 (93.3)	662 (91.8)
Missing	1 (0.2)	0 (0)	0 (0)	1 (0.1)

BW, birth weight; GA, gestational age; PMA, postmenstrual age.

^aResults given as number (%).

^bFisher exact test.

Overall, 691 infants (95.8%) received outpatient follow-up, with only 30 (4.2%) completely lost to follow-up. For those without follow-up, reasons were largely unknown (n = 21 [70%]). Documented reasons included patient illness (n = 3), misunderstanding the need/expectations for ROP follow-up care (n = 2), another family member needing medical care (n = 1), transportation difficulties (n = 1), the parent being busy or overwhelmed (n = 1), or being unable to be contacted altogether (n = 1). With some characteristics (language and ROP stage for LTFU, ROP stage for delays), logistic regression could not be performed due to excessively small group size. Compared to zone II ROP, zone III ROP or retinal maturity at the last inpatient examination was associated with complete LTFU (OR = 0.20; 95% CI, 0.09-0.45; P <0.001). Infants ≥40 weeks’ PMA at time of discharge also were more likely to be LTFU (OR = 0.31; 95% CI, 0.14-0.65; P = 0.002) (Figure 1).

FIG 1.

Association of factors with any ROP follow-up.

Of those with follow-up (n = 691), 89 (13%) were delayed by a median of 9 days (range, 4-86) from the recommended date. Similarly, the cause of most delays could not be identified through chart review (n = 48 [54%]). Of reasons documented, however, the most commonly identified issue was transportation (n = 16 [39%]), followed by patient illness (n = 15 [37%]), illness of family or friends (n = 2 [4.9%]), caregivers being busy or overwhelmed (n = 2 [4.9%]), misunderstanding the need/expectations for ROP follow-up care (n = 2 [4.9%]), weather (n = 2 [4.9%]), financial concerns (n = 1 [2.4%]), and difficulty contacting caregivers (n = 1 [2.4%]).

Using a binary categorization of race/ethnicity, Hispanic and/or non-White infants had significantly lower odds of timely follow-up (OR = 0.44; 95% CI, 0.28-0.69; P < 0.001) in univariate logistic regression. Inpatient ROP treatment was associated with greater odds of timely follow-up (OR = 5.27; 95% CI, 1.11-94.2; P = 0.03). Among patients who had any follow-up, those with PMA of 40 weeks or more at discharge had higher odds of timely follow-up (OR = 1.84; 95% CI, 1.08-3.30; P = 0.03) (Figure 2).

FIG 2.

Association of factors with timely ROP follow-up.

Results of the Poisson regression sensitivity analysis were broadly consistent with the main logistic analysis, in terms of which variables met statistical significance and the direction of effects, though with differences in the estimated effect sizes and confidence intervals (Supplement 2, available at jaapos.org, Tables S1-S2).

The P values for Fisher exact tests are shown in Table 2. Consistent with the regression results, ROP zone (P = 0.002), PMA (P = 0.002), and race/ethnicity (P = 0.02) were significantly associated with follow-up status.

Discussion

The primary objective of this study was to determine the rate and determinants of delayed or missed follow-up for the first outpatient ROP examination after hospital discharge. Our study findings suggest that a majority of infants are adherent to outpatient ROP screening follow-up, though delayed follow-up remains a significant challenge; that Hispanic and/or non-White patients may be less likely to have timely follow-up; and that patients who have more severe ROP (as indicated by treatment or worse zone) may be more likely to adhere to follow-up recommendations.

The rate of follow-up for this cohort (96%) was higher than that for smaller single-institution studies, with adherence rates between 42%-81%, but comparable to larger multicenter studies, with 94% adherence.^9,11,12 However, a considerable number of follow-ups (13%) were delayed. To our knowledge, this is the first large-scale study to also analyze rates and predictors of delayed and nondelayed ROP follow-up. As such, this study presents a more nuanced understanding of the barriers not just to overall adherence but to timely care, which is clinically relevant given the poor visual outcomes associated with untimely management.⁴

Maternal race, insurance status, and social deprivation, as well as clinical factors of low birth weight and ROP treatment status, have all been associated with ROP follow-up adherence.^8,13,14 Our findings reinforce the existing body of literature by highlighting the role of sociodemographic and clinical factors in ROP follow-up and extend this analysis to the timeliness of care.

In univariate analyses, timely follow-up was significantly more likely for non-Hispanic White infants and those who had received ROP treatment while inpatient, consistent with the reported predictive value of these factors for overall ROP adherence in previous literature.¹³ Unfortunately, we were unable to perform multivariable analysis to control for potential confounding because of the limitations of this particular dataset. Our study is therefore exploratory, and the results must be interpreted with caution. That said, there are reasons to believe that the characteristics we identified could have a real effect. For example, families with infants who have more severe, treatment-requiring disease may receive more in-depth counseling throughout their admission and at the time of discharge, potentially leading to better follow-up.

Race and ethnicity are social constructs without biological basis, but they present crucial implications for what may be closely related predictors that could mediate the apparent effect of race on follow-up adherence, such as socioeconomic status (SES). Though SDI was not a significant predictor of delayed follow-up in our univariate analysis, it is not a comprehensive measure of SES.¹⁰ The impacts of race on SES may go beyond SDI to include factors like occupation, wealth, and social networks, which are more difficult to quantify and were not directly assessed in this study.¹⁵

Higher ROP zone in the worse-seeing eye was significantly associated with complete LTFU in this cohort. This may be due to a lower sense of caregiver urgency when ROP is less severe; it may also reflect physicians’ clinical judgment in determining how beneficial it is to escalate follow-up efforts for low-risk patients in specific situations.

There was also an association between higher PMA at time of discharge and complete LTFU. Paradoxically, among those with high PMA who had any follow-up, there was a higher chance of timeliness. It is not fully clear whether this reflects factors that could impact caregiver capacity or perceived urgency of follow-up (eg, admission duration, number of inpatient screenings already undergone, overall health status), or if this is a pattern specific to this dataset. While not impossible, these paradoxical results raise the question of a spurious finding. It would be informative to see if this is replicated in future studies, augmented with more robust qualitative data.

Encouragingly, transportation was not often reported as an issue for those with complete LTFU (n = 1). We suspect that this reflects caregivers being connected with social work support once a need was identified, but there may be a lag, because transportation was still a reported barrier for timely follow-up. There was no significant association between driving time and delayed follow-up, indicating that the presence of transportation rather than the physical distance has greater impact on follow-up. These qualitative findings reveal specific caregiver concerns that should be addressed, especially as prior studies have shown that over half of canceled or missed appointments are for caregiver reasons.⁶

Policy changes, such as streamlined scheduling and attendance logs, in combination with parent education have also been shown to increase first ROP screening attendance rates.^11,12 Other studies have referenced earlier involvement with CPS to connect caregivers with resources for earlier intervention, though there is risk of unintended negative social consequences.⁸ One opportunity for policy change could be through the ROP Toolkit, freely provided by the Ophthalmic Mutual Insurance Company (https://www.omic.com/resource/3rd-edition-of-rop-creating-a-safety-net-published/), a medical malpractice insurance company endorsed by the American Academy of Ophthalmology. The addition of a discharge screening questionnaire with items regarding barriers to timely care (eg, transportation or childcare support) could be a tangible method of connecting caretakers with resources, possibly preventing downstream delays in care. More research regarding how these toolkits are used by institutions and providers would elucidate the potential impact of this change. Ultimately, early policy change is prudent in this area, as emerging research predicts a workforce shortage in professionals involved in the care of ROP.¹⁶

This study has important strengths and limitations. Its main strength is the large sample size and years of data collected which improves external validity. Still, as a retrospective study at a single institution, generalizability is limited. For statistical purposes, we aggregated socioeconomic and racial and ethnic groups, which does not accurately represent the spectrum of sociodemographics and how they interact with each other. Similarly, SDI was assessed at the level of zip code, which may have led to loss of information compared to individual-level measures. Though another strength of the study is the inclusion of qualitative data, reasons for delay or LTFU were largely unavailable, limiting generalizability of the findings. The lack of process standardization and dependency on human resources may skew which types of reasons are documented. Quality improvement measures may be considered to standardize the documentation process and enhance research with caregiver perspectives. As mentioned earlier, the inability to control for confounders is a significant limitation, but these exploratory findings may still provide useful information to guide future work.

In conclusion, while this study demonstrates generally successful follow-up rates, overall adherence is only clinically meaningful when ROP care is appropriately timely. The persistence of racial and ethnic disparities in delays to care highlights the need for targeted interventions. Transportation assistance may be of benefit to caregivers, although standardized documentation systems would provide a more comprehensive understanding of caregiver-related barriers.