Risk Factors and Outcomes of Delayed Presentation of Diabetic Retinopathy Patients to a County Hospital

Abstract

Purpose: To identify risk factors and evaluate outcomes of patients with delayed presentation and advanced diabetic retinopathy in our safety-net county hospital population.

Methods: A retrospective study was performed on 562 patients who presented with a new diagnosis of diabetic retinopathy (DR). Delayed presentation was defined as moderate or severe nonproliferative diabetic retinopathy (NPDR) or proliferative diabetic retinopathy (PDR) at the initial visit. Comparisons between patient groups were performed with chi-square or Fisher's exact test for categorical variables and multinomial logistic regression for multivariable analysis. Linear and logistic regression modeling with general estimating equations to account for patients having two eyes was used to compare eye-level outcomes.

Results: Lack of a primary care provider (PCP) was highest in patients who presented initially with PDR (28.8%), compared to 14.3% in moderate/severe NPDR, 12.4% in mild NPDR, and 7.6% in no DR groups (P < 0.001). Only 69.4% of patients with a PCP had an ophthalmology screening referral. Highest lack of referral (47.2%) was seen in the PDR group (P = 0.002). Patients with PDR were more likely to be uninsured (19.2%) compared to no and mild DR groups, with rates of 7.6% and 9.0%, respectively (P = 0.001). The PDR group had worse initial and final visual acuities (P < 0.001).

Conclusions: Several risk factors were noted for delayed DR presentation, including lack of PCP, lack of screening referral, and uninsured/underinsured status. Patients with advanced DR at presentation had worse final visual outcomes despite aggressive treatment.

Translational Relevance: Screening programs targeting populations with identified risk factors are essential for improving outcomes.

Introduction

Globally, the prevalence of diabetes and its impact continues to increase, especially in low- and middle-income communities.¹ Recent projections estimate adults with diabetic retinopathy (DR) and vision-threatening diabetic retinopathy will increase from 103.12 million and 28.54 million in 2020 to 160.50 million and 44.82 million by 2045, respectively.² In the United States, DR is projected to rise in adults 40 years and older, with projections of 5.5 million in 2005 to 16 million in 2050.³ This poses a serious public health issue as DR is the leading cause of new-onset blindness in American adults 24 to 70 years old.⁴ These increased rates will lead to substantial rises in medical costs, lost productivity, and decreased quality of life. 

Policies aimed at early intervention are well supported by the literature. Diabetes duration and poor glycemic control are two significant risk factors for retinopathy. Intensive glycemic control reduces risk of diabetic retinopathy development and progression.⁵ The timing of treatment is also important. Randomized clinical trials have shown early treatment of proliferative diabetic retinopathy (PDR) can reduce the rate of blindness.⁶ Therefore, ophthalmology care early in the disease course is critical. However, early presentation is not often the norm, and lack of proper diabetic retinopathy screening (DRS) is most prevalent in lower socioeconomic groups.⁷ The rate of diabetic retinopathy development is also higher in lower socioeconomic status patients and occurs earlier than in those in higher socioeconomic groups.⁸ Many factors contribute to difficulty in early DRS. Patients with DR have suboptimal adherence to eye care appointments.⁹ Unstable housing corresponded to an over fivefold risk of diabetes-related emergency department visits or hospitalizations in a cross-sectional survey of patients receiving care at safety-net health centers in the United States.¹⁰ Even among insured patients, the cost of insurance deductibles can be the difference between seeking or delaying care. Wharam et al.¹¹ analyzed the impact on diabetic adults of an employer-mandated switch from a low- to high-deductible health plan and found a 4% decline in total health care expenditures and an approximately 25% increase in high-severity emergency department visits and hospitalizations in members who lived in low-income neighborhoods. 

Understanding risk factors for delayed presentation is the first step toward implementing strategies that facilitate early referral and care for at-risk populations. In this study, we aimed to investigate risk factors for delayed presentation of diabetic retinopathy in a safety-net hospital in Denver, Colorado, USA. 

Methods

This retrospective cohort study was approved by the Colorado Multiple Institutional Review Board (IRB), and participant consent was not required due to IRB waiver of consent. This study was carried out in accordance with the principles of the Declaration of Helsinki. Patients seen at the Denver Health Eye Clinic at Denver Health Medical Center (DHMC) in Denver, Colorado, were eligible for study inclusion. DHMC is Colorado's primary safety-net institution with an urban public hospital and community health centers that care for uninsured and other vulnerable populations. Between January 2014 and December 2020, patients who were seen with any DR stage were identified by all DR diagnosis codes through an electronic medical record (EMR) tool for study inclusion. Inclusion criteria were a complete EMR and DR treatment-naive status. Exclusion criteria were deceased status, significant other comorbid eye conditions, and correctional facility status. 

The EMR review and data collection in RedCap were conducted on patients who met study criteria. Importantly, the following data, along with baseline characteristics and clinical information, were extracted: (1) DR severity at initial and final visits, (2) appointment adherence, (3) referral source, and (4) DR treatment received. Health insurance status was also collected, including two safety-net programs specific to Colorado, Colorado Indigent Care Program (CICP) and Denver Financial Assistance Program (DFAP). CICP is a statewide program providing discounted health care services for low-income patients ineligible for Medicaid. DFAP is a Denver Health–specific program that provides discounted health care services for low-income patients not eligible for Medicaid or CICP. 

Patient demographic variables included in the analysis were age, gender, ethnicity, insurance type, and primary care provider (PCP) status. Other characteristics, such as diabetes duration, primary language, comorbid conditions, and most recent HbA1c, were captured at the last visit. 

The primary outcome measures were change in DR status from initial to final visit and visual acuity (VA), defined in logarithm of the minimum angle of resolution (logMAR) values, from initial to final visit. DR severity was individually evaluated by each patient's optometrist or ophthalmologist, with five recognized stages that follow the International Clinical Disease Severity Scale¹² for DR: (1) no retinopathy, (2) mild nonproliferative retinopathy (NPDR), (3) moderate NPDR, (4) severe NPDR, and (5) PDR. Delayed presentation was defined as initial presentation with moderate NPDR, severe NPDR, or PDR. Early presentation was defined as presenting with either no diabetic retinopathy or mild DR. 

Definition of Adherence Indicators

Previously defined intervals from a study by Bresnick et al.⁹ on adherence to ophthalmology appointments were used in this investigation of appointment adherence status. Three separate measures of adherence were determined for patients with a PCP, and adherence was categorized as follows: 

• 1. Nonadherent: no follow-up visits after initial appointment
• 2. Partially adherent: returned for some follow-up visits but failed to return for one or more visits within 6 months of the recommended follow-up interval
• 3. Completely adherent: returned for all follow-visits within 6 months of the recommended intervals
• 4. Indeterminate: could not determine adherence due to missing data or inadequate chart review interval

Inadequate chart review interval was defined as lack of sufficient time (6 months) between chart review and last recommended visit date. 

Statistical Analysis

Descriptive statistics are described with basic frequencies, means, and standard deviations. Analyses were performed at both patient and eye level. Patient-level analyses were performed on characteristics deemed specific to the patient, such as primary language, insurance status, and adherence. For patient-level analyses, patients were categorized into four groups based on the most advanced eye if severities were not symmetric: no DR, mild NPDR, moderate and severe NPDR, or PDR. The moderate and severe NPDR groups were combined due to small sample size in the severe NPDR category. 

Analyses of DR progression and DR treatment were analyzed at the eye level, as patients’ eyes may have differed regarding these factors. Patients lacking a final visit were excluded from VA outcome and DR status change analyses as we could not measure their final visit outcomes or progression. Recorded Snellen VAs were converted to logMAR values for the purpose of statistical analysis. 

Comparisons between patient groups were performed with the chi-square or Fisher's exact test for categorical variables and analysis of variance, Kruskal–Wallis, or Wilcoxon rank-sum test for continuous variables. Multinomial logistic regression modeling was used for multivariable analysis that included sex, age, and variables with P < 0.10 in univariate analysis and few missing values. This model is presented so as to include all patients. Since some variables had a lot of records with missing values, a sensitivity analysis of the multivariable model was performed among patients with no missing data that included all variables with P < 0.10 in univariate analysis (despite having missing data). For this sensitivity analysis, adherence and having a PCP were combined into one variable since adherence was only designated for patients who had a PCP. Linear and logistic regression modeling with general estimating equations to account for patients having two eyes included were used to compare eye-level outcomes. A P value less than 0.05 was considered significant for all statistical comparisons. Statistical analyses were performed using SAS version 9.4 statistical software (SAS Institute, Cary, NC, USA). 

Results

Of the 830 patients who met study inclusion criteria, 103 were excluded for previous treatment, 15 for significant comorbid eye conditions, 16 for correctional facility status, and 134 for deceased status during study period (n = 268). From our review, we found 562 DR treatment-naive patients seen for DR between January 2014 and December 2020 who met inclusion criteria. Baseline characteristics are shown below, with a mean age of 56.5 ± 11.4 years and the majority (60.8%) identifying as Hispanic (Table 1). Most (70.3%) had treatment for hypertension. Among the 527 patients with an HbA1c in their EMR, the average of their most recent was 8.5 ± 2.0, ranging from 4.6 to 14.2. 

Table 1.

 

Baseline Characteristics of Study Participants (Initial Visit) (N = 562)

 

Table 2 shows comparison of patient characteristics by DR severity at initial visit. Spanish as a primary language was more prevalent in those who presented with PDR versus no retinopathy (58.6% vs. 33.8%). In addition, 71.2% of patients with PDR had a PCP compared to 92.4% of the no retinopathy group and 87.6% of the mild NPDR group. Of patients with a PCP, 69.4% (327/471) had a referral while those without a PCP had a referral rate of 48.9% (44/90; P < 0.001). Highest lack of referral was seen in the PDR group (47.2%). Only 18.8% of the PDR group were fully compliant with PCP appointments compared to 40.9% in the no DR group and 36.8% of the mild NPDR group. Uninsured status was highest in the PDR group at 19.2%. Higher HbA1cs, with means of 8.7, were found in both the moderate/severe NPDR and PDR groups, compared to the no DR group with a mean A1c of 8.2 (P = 0.03 and P = 0.04, respectively). Finally, the rate of kidney disease was highest in the PDR group at 49.6%. The final multivariable model included sex, age, primary language, kidney disease, having a PCP, and insurance status. Primary language, kidney disease, having a PCP, and insurance status all remained significant. In the sensitivity analysis that also included duration of diabetes, HbA1c, and a combined variable of having a PCP and adherence with PCP, the variables of primary language, kidney disease, and insurance status remained significant. Sex, age, duration of diabetes, HbA1c, and the combined variable of having a PCP and adherence were not significant. 

Table 2.

 

Characteristics of Patients by Severity of Diabetic Retinopathy at Initial Visit

 

Eye-level data of outcomes for the 1,124 eyes at final visit are presented in Table 3 by DR status of that specific eye at initial visit. Within the advanced DR eyes, a high portion of eyes with PDR received treatment, with 77.9% receiving panretinal photocoagulation (PRP), 60.4% receiving injections, 34.2% receiving pars plana vitrectomy (PPV), and 27.5% receiving cataract surgery. The moderate/severe NPDR eyes received less treatment but still at higher rates than the early presentation group, with 17.2% receiving PRP, 26.2% receiving injections, 2.3% receiving PPV, and 20.4% receiving cataract surgery. In comparison, within the eyes with early presentation, the no DR and mild DR eyes had 4.1% and 3.6% receiving PRP, 7.6% and 4.4% receiving injections, 2.9% and 0.5% receiving PPV, and 20.6% and 16.0% receiving cataract surgery, respectively. 

Table 3.

 

Initial and Final Visual Acuity in Early and Delayed Presentation of Diabetic Retinopathy

 

The mean follow-up time in months from initial to final visit was longest for patients with no DR, 33.8 ± 14.4, and shortest for patients with PDR, 17.9 ± 14.8, and was significantly different across all four groups (P < 0.0001). The mild NPDR group's mean follow-up was 24.8 ± 15.8, and the moderate/severe group was 19.8 ± 16.5. At initial visit, the mean VAs in logMAR values for the delayed presentation groups of moderate/severe NPDR and PDR were 0.40 ± 0.44 (20/50) and 0.96 ± 0.82 (20/182), respectively, in comparison to the early presentation groups of no DR and mild NPDR with VAs of 0.36 ± 0.51 (20/46) and 0.31 ± 0.38 (20/41) (P < 0.001). There was no significant difference across groups in VA decline from initial to final visit (P = 0.20). However, PDR eyes had the worst mean final VA at 0.87 ± 0.80 (20/148), while the no DR, mild NPDR, and moderate/severe NPDR groups had mean VAs of 0.33 ± 0.46 (20/43), 0.31 ± 0.46 (20/41), and 0.36 ± 0.42 (20/46), respectively (P < 0.001; Table 3). 

Discussion

To the best of our knowledge, this is the first study to evaluate multiple risk factors for patients with delayed presentation of DR at a county hospital in the United States. Over half of the patients in the study presented with moderate/severe NPDR or PDR, as we defined as delayed presentation, emphasizing the need to identify risk factors for this delay. One study, a 2004 US hospital-based pilot case control study, found an association between socioeconomic status and late DR presentation but had a limited sample size (N = 52).¹³ In our study, we had a substantially larger cohort and identified multiple risk factors for delayed DR presentation. 

Patients with Spanish as a primary language had higher rates of delayed DR presentation, compared to English as a primary language. All non-English-speaking patients had the option of in-person or telephone interpretation. Language barriers have been found to negatively impact health care delivery and patient satisfaction. Al Shamsi et al.¹⁴ found patients may avoid seeking eye care earlier due to this language barrier. Clinical explanations can become lost in translation, and interpretation may not always be efficient. Many patients who do not speak the local language as their primary language face issues understanding medical situations and medication instructions.¹⁵ These issues can contribute to non-English-speaking patients not seeking health care as frequently as those with English as their primary language.¹⁴ 

Patients who presented with advanced DR were approximately twice as likely to be uninsured or underinsured with DFAP and CICP than those who presented with early DR. Lack of health insurance has been established as a major obstacle to eye care.¹⁶ Uninsured or underinsured patients seek care at suboptimal rates due to fear of out-of-pocket medical costs. Based on our findings, DRS programs targeting uninsured or underinsured patients may be helpful in reducing the rate of patients presenting with late-stage DR. 

However, even with insurance, Benoit et al.¹⁷ demonstrated the percentage of eye exams meeting American Diabetes Association (ADA) recommendations for DRS was below 30%. In addition, even when barriers such as cost and accessibility were removed in a study by Keenum et al.,¹⁸ patient adherence to appointments within indicated time frames was low—only 30% of patients were compliant within recommended guidelines. In our study, poor compliance with PCP appointments was a risk factor for delayed presentation. PCP compliance was lowest in patients who presented with advanced DR. This likely reflects overall barriers to medical care, such as lack of transportation, lack of health care literacy, and inability to take time off work for appointments.¹⁹ 

Urgent visits were seen at the highest rate in the PDR group, as patients likely did not seek care until a vision-threatening event occurred. Unfortunately, the advanced DR presentation group had worse visual acuities at initial visit, and while all groups experienced similar rates of visual decline, their visual acuities remained lowest at final visit. This occurred despite treatment (PRP, injections, surgery) and quick referral-to-visit times. This emphasizes that early presentation is crucial for vision preservation. 

Our study also found delayed presentation patients were less likely to have a PCP. However, despite having a PCP, 30.6% of patients with a PCP lacked a diabetic eye examination referral. PCPs have cited barriers to referral, including scheduling difficulties and “poor communication” from eye care providers.²⁰ It has been found there is little oversight from PCPs on whether patients attended their eye appointments.²¹ These factors contribute to the fact that fewer diabetic patients are being screened than would be expected based on ADA recommendations—less than half of diabetic patients in a study by Pérez et al.²² received an annual dilated eye examination. 

Greater DRS initiatives are needed to target populations most vulnerable to nonadherence for eye care. Mobile eye screening clinics have increased examination rates among these populations, likely by alleviating the transportation barrier and, often, the cost barrier. Relevantly, Chheda et al.²³ found many patients screened by the mobile eye clinic who were non-English speaking and were uninsured had no previous eye exam—indicating these clinics could be useful in targeting patients with two of the risk factors for delayed DR presentation that we identified. While mobile clinic studies have been conducted in rural areas, there could be benefits to implementing such a program around Denver and other metropolitan cities. 

In addition, there is recent evidence that telemedicine DRS programs in primary care settings have improved DR screening rates, especially among patient groups that most need referral to eye care.^24,25 This is likely due to removal of barriers to care, such as transportation and clinic wait times. Given the significant rise in telemedicine over the past year,²⁶ clinicians are likely to continue utilization of virtual screenings, which would be of great benefit to patients most vulnerable to DR progression. 

There are limitations to this study. This investigation was conducted in an eye clinic of a large metropolitan county hospital, serving a mostly low-income, ethnically diverse population with primarily Medicaid coverage. Our study findings, therefore, are most likely to be applicable to similar settings and may have limited generalizability in higher-income populations. In addition, while the overall sample size was high compared to prior studies related to this subject, the severe NPDR group was relatively small at 40 patients and was combined with the moderate NPDR group for the purpose of statistical analyses. Ideally, the follow-up time from initial to final visit would have been longer. Finally, retrospective chart review is limited by being subject to confounding that may not be captured, difficulty assessing temporal relationships, and often having missing data, which was the case in our study cohort. 

In summary, lack of a PCP, lack of diabetic eye examination referral, and uninsured status are significant risk factors associated with presenting with advanced DR in our safety-net hospital population. This, in turn, is associated with worse visual outcomes despite aggressive treatment. Our findings emphasize the need for diabetic retinopathy screening programs targeting these at-risk patient populations. 

Acknowledgments

Supported by a grant from Research to Prevent Blindness to the Department of Ophthalmology. 

Disclosure: V.I. Lu, None; J.L. Patnaik, None; R.A. Scott, None; A.M. Lynch, None; J.M. Smith, None; N. Mandava, Soma Logic (C, F), ONL Therapeutics (C), Alcon (P), 2C Tech (P, O), Aurea Medical (O); N. Manoharan, Iveric Bio (F), Genentech (F)