Patients Presenting Emergently with Proliferative Diabetic Retinopathy: Follow-up and Factors Associated with Compliance

Abstract

Purpose:

To determine the rate of follow-up after emergent encounters for Proliferative Diabetic Retinopathy (PDR) and to identify patient or visit characteristics associated with follow-up compliance.

Methods:

A retrospective cohort study of patients presenting to an ophthalmic Emergency Department (ED) with active PDR between May 2014 and December 2016 was conducted. Demographic data and encounter data were gathered for each ED visit. Compliance with follow-up was defined as a completed clinic visit as scheduled after the emergency encounter.

Results:

A total of 590 ED encounters were included. The overall follow-up rate was 61.9%. Married patients and those with Public Health Trust insurance had increased odds of compliance (OR: 1.507, p = 0.04; OR: 2.749, p < 0.0001). Patients with Medicaid had reduced odds (OR: 0.543, p = 0.004). Patients with longer ED encounters and longer intervals to follow-up had reduced odds (OR: 0.948, p = 0.001; OR of 0.941, p < 0.0001). The other characteristics were not significantly associated with follow-up compliance.

Conclusions:

Patients who present emergently with active PDR are at high risk for following up non-compliance. Characteristics with significant effects on the odds of follow-up compliance include specific insurance payer, marriage status, length of visit, and interval to follow-up.

Summary Statement.

Patients presenting emergently with proliferative diabetic retinopathy have low rates of follow-up compliance, 61.9%. This trend is consistent across nearly all demographic variables. The specific insurance payer, length of visit, and interval to follow-up were found to have a significant impact on follow-up compliance and should help identify high risk patients.

INTRODUCTION

Diabetes and its complications are an evolving public health crisis. The Center for Disease Control (CDC) estimates that 30.3 million people in the United States had diabetes in 2015, representing nearly 1 in 10 Americans.¹ Even conservative assessments predict that the prevalence of diabetes in the US will increase by nearly 20 million by 2030.² Treating the complications of diabetes accounts for a significant and increasing portion of total health expenditures, with some estimates topping $100 billion dollars each year.³ The burden will continue to increase unless more effective measures are taken to address preventable complications of diabetes.^4,5

One of the morbid consequences of diabetes is diabetic retinopathy. This microvascular complication is a major cause of vision loss and has been estimated to occur in 28% of diabetic patients over 40, with rates approaching 40% in subsets of the population.⁶ Proliferative diabetic retinopathy (PDR) represents advanced retinal disease, characterized by neovascularization of the optic nerve head and retina. Effective interventions exist to reduce progression beyond the non-proliferative stages, and there has been demonstrable success in decreasing the burden of PDR.^7,8 In cases where secondary prevention fails, available treatments effectively decrease the risk of severe vision loss associated with PDR.^9,10 Nevertheless, medical, laser, or surgical options all require long term care with consistent follow-up to be successful.

In spite of efficacious interventions, PDR has been estimated to affect 1.5% of adults with diabetes in the United States. Nearly half of patients with high risk characteristics are expected to experience profound visual loss without treatment.^11,12 Patients with PDR may experience sudden vision loss from complications including macular edema, vitreous hemorrhage, or tractional retinal detachment, and these changes often prompt emergency medical evaluations. Ideally, these visits would result in subsequent treatment for this at-risk population. However, the frequency and predictors of continued care for emergent PDR patients have not been reported in the literature to date. The current study aimed to determine the rate of follow-up after ED encounters for PDR and to identify factors associated with follow-up compliance.

METHODS

The approval of the University of Miami Institutional Review Board was obtained prior to conducting this study, which was performed in accordance with the Health Insurance Portability and Accountability Act of 1996 and adhered to the tenets of the Declaration of Helsinki. The setting for this study was a single institution, the Bascom Palmer Eye Institute (BPEI). A database search of the Emergency Department (ED) encounters between May 2014 and December 2016 was performed. All encounters for PDR and its complications were identified in the electronic medical record (EMR) using ICD 9 and ICD 10 codes. The EMR for each of these encounters was manually reviewed retrospectively to ensure all patients had active PDR, indicated by neovascularization of the optic nerve disc, neovascularization of the retina, or vitreous hemorrhage. Patients in whom PDR was an inactive problem at the time of the encounter were excluded. For those patients with active PDR, demographic data and encounter data were collected based on the standardized information provided by all ED patients at intake. Demographic data included sex, ethnicity, race, primary language, age, marital status, and address. The zip code of this address was used to determine the distance to BPEI using Google Maps.¹³ The zip code was also used to determine median household income using data from the US Census Bureau 2012 – 2016 American Community Survey 5-Year Estimates.¹⁴ Recorded data specific to the emergent encounter included insurance payer, visual acuity (VA) in the right and left eye, length of encounter, and emergent treatment (either panretinal laser photocoagulation (PRP), or intravitreal injection (IVI) of an anti-vascular endothelial growth factor (VEGF) agent, or both).

The discharge protocol for each patient involved making a follow-up appointment as specified by the ED physician. For the purpose of this study, the follow-up appointment was defined as the next encounter with a BPEI provider scheduled in the EMR. If no follow-up encounter was scheduled despite the physician ordering one, this was recorded. Patients documented in the EMR as following up with an outside provider were excluded because no follow-up data was available. If a patient returned to the ED prior to their next scheduled appointment, this ED visit was recorded as the follow-up appointment. The status for the follow-up appointment was documented as completed, canceled, or no show as recorded in the EMR. Follow-up compliance was defined as a completed encounter within the time frame ordered by the ED physician. Follow-up non-compliance included those follow-up appointments recorded as canceled, as no-show, as not scheduled, as not completed within the period ordered by the physician, or when the immediate follow-up was another encounter in the ED.

Chi square goodness-of-fit tests were used to assess whether demographic characteristics of the patients (age, sex, ethnicity, race, household income) differed from the general population of Miami-Dade County as reported by the United States Census Bureau in its 2012–2016 estimates.¹⁵ Logistic regression was used to estimate the odds ratio (OR) for an association between possible explanatory variables and the outcome variable of follow-up compliance using SAS version 9.4 (Cary, NC) PROC GENMOD, which accounted for the correlation between the two eyes of each patient. For ethnicity, the reference category was non-Hispanic. For race, the reference category was Black or African-American, which was compared to White/Caucasian and to Other/Multiracial. This last group included anyone coded as Other, Multiracial, Asian, or American Indian or Alaskan native. For primary language, the reference category was English. This was compared to Spanish, to Creole, and to Other (a combined category that included Tamil, Arabic, and unknown). For marital status, the reference category was single. This was compared to married, to divorced/separated, and to widowed. Some patients declined to answer questions regarding ethnicity, race, primary language, or marital status. For payer, insurance providers were categorized as Commercial (including Preferred Provider Organizations and Health Maintenance Organizations), Medicare, Medicaid, Public Health Trust (PHT, a publicly funded insurance for low-income residents of MDC), or Self Pay. There was no reference category for payer. Instead, each payer was compared to the average of all other payers. For treatment, the reference category was no treatment, which was compared to any treatment (PRP, IVI, or both). For continuous explanatory variables, the ORs were calculated for a 10 year increase in age, a $5000 increase in the median neighborhood household income, an increase of 5 miles in the distance to BPEI, a 1-unit increase in logMAR VA, a half-hour increase in the length of the initial encounter, and a 1-week increase in the time to follow-up.

RESULTS

The mean patient age for all encounters was 55.6 (Standard Deviation (SD) 11.5) years and the majority of the encounters were for male (55%) and Hispanic (53%) patients. When asked about race, 46% identified as White/Caucasian compared to 35% as Black/African American or 16% as Other/Multiracial. The majority of patients reported English as their primary language (55%), with Spanish (38%) and Creole (6.4%) comprising smaller portions of the total. Patients most commonly reported being married (41%) or single (39%). The median for the median household income by zip code was $39,386 (Interquartile Range (IQR) $19,000). Patient demographic characteristics are summarized in Table 1. Compared to Miami Dade County (MDC), the population of PDR patients was significantly different. There was a higher percentage of male patients (PDR 55%, MDC 49%, p = 0.002), non-Hispanic patients (PDR 45%, MDC 34%, p <0.0001), Black/African American patients (PDR 35% v 19%, p < 0.001), patients 65 years or older (PDR 20%, MDC 15%, p = 0.02), and patients with household incomes less than $44,224 (PDR 73%, MDC 50%, p < 0.001).

Table 1.

Demographic Characteristics of Emergent Proliferative Diabetic Retinopathy Patients

	Total	Follow-up Non-compliant	Follow-up Compliant
Gender, n (%)
Male	324 (55%)	119 (37%)	205 (63%)
Female	266 (45%)	106 (40%)	160 (60%)
Age (years), n (%)
<40	58 (9.8%)	22 (38%)	36 (62%)
40-49	95 (16%)	32 (34%)	63 (66%)
50-59	220 (37%)	88 (40%)	132 (60%)
60-69	160 (27%)	62 (39%)	98 (61%)
>70	57 (9.7%)	21 (37%)	36 (63%)
Ethnicity, n (%)*
Non-Hispanic	268 (45%)	111 (41%)	157 (59%)
Hispanic	311 (53%)	113 (36%)	198 (64%)
Race, n (%)*
Black/African-American	209 (35%)	85 (41%)	124 (59%)
White/Caucasian	272 (46%)	98 (36%)	174 (64%)
Other/Multiracial	95 (16%)	40 (42%)	55 (58%)
Primary Language, n (%)*
English	327 (55%)	124 (38%)	203 (62%)
Spanish	223 (38%)	88 (39%)	135 (61%)
Creole	38 (6.4%)	12 (32%)	26 (68%)
Other	2 (0.3%)	1 (50%)	1 (50%)
Marital Status, n (%)*
Single	228 (39%)	95 (42%)	133 (58%)
Married	241 (41%)	78 (32%)	163 (68%)
Divorced/Separated	91 (15%)	36 (40%)	55 (60%)
Widowed	27 (4.6%)	13 (48%)	14 (52%)
Median Household Income
Mean (SD)	$39,658 ($15,000)	$39,108 ($14,000)	$39,997 ($15,000)
Median (IQR)	$39,386 ($19,000)	$39,007 ($18,000)	$39,386 ($21,000)

^*remaining patients in these categories declined to answer

Visual acuities at presentation were generally poor. The median acuity for the better seeing eye was 0.48 (IQR 0.8) LogMAR (20/60 Snellen Equivalent) and for the worse seeing eye was 1.6 (IQR 1.4) LogMAR (20/800 Snellen Equivalent). Medicare was the most common payer (32%), followed by commercial insurance (HMO/PPO 28%), Medicaid (21%), Public Health Trust (16%), and self-pay (3.4%). A small number of patients received treatment in the ER (19%), with PRP alone being the most common modality. The median distance between home zip code and the clinic was 11.3 (IQR 14) miles. The median duration of the initial ED encounter was 169 (IQR 119) minutes and the median for interval until the next encounter was 9 (IQR 14) days. Encounter characteristics are summarized in Table 2.

Table 2.

Encounter Characteristics of Emergent Proliferative Diabetic Retinopathy Patients

	Total	Follow-up Non-compliant	Follow-up Compliant
Payer, n (%)
Medicare	190 (32%)	71 (37%)	119 (63%)
Commercial (PPO/HMO)	164 (28%)	66 (40%)	98 (40%)
Medicaid	122 (21%)	61 (50%)	61 (50%)
Public Health Trust	93 (16%)	19 (20%)	74 (80%)
Self-Pay	20 (3.4%)	7 (35%)	13 (65%)
Other	1 (0.2%)	1 (100%)	0 (0%)
Treatment in ED, n (%)
No Treatment	479 (81%)	183 (38%)	296 (62%)
Treatment	111 (19%)	42 (38%)	69 (62%)
PRP only	64 (11%)	24 (37.5%)	40 (62.5%)
IVI only	43 (7.3%)	18 (42%)	25 (58%)
Both PRP and IVI	4 (0.8%)	0 (0%)	4 (100%)
Distance to Clinic (miles)
Mean (SD)	39.3 (165)	38.1 (169)	40.1 (163)
Median (IQR)	11.3 (14)	11.1 (11)	11.9 (16)
Better Eye Acuity (logMAR, Snellen)
Mean (SD)	0.63 (0.6), 20/85	0.59 (0.6), 20/78	0.66 (0.6), 20/91
Median (IQR)	0.48 (0.8), 20/60	0.40 (0.5), 20/50	0.48 (0.8), 20/60
Worse Eye Acuity (logMAR, Snellen)
Mean (SD)	1.52 (0.8), 20/662	1.45 (0.8), 20/564	1.55 (0.8), 20/710
Median (IQR)	1.60 (1.4), 20/800	1.60 (1.7), 20/800	1.60 (1.5), 20/800
ED Visit Length (minutes)
Mean (SD)	198 (105)	213 (111)	189 (99.6)
Median (IQR)	169 (119)	181 (112)	163 (119)
Interval to Next Encounter (days)
Mean (SD)	17.9 (57)	32.4 (92)	9.5 (8.1)
Median (IQR)	9 (14)	15 (20)	7 (11)

Inclusion criteria were met by 590 emergent encounters, which represented 429 unique patients. The overall rate of follow-up after each encounter was 61.9%. For those patients who did not follow-up, the majority of the appointments were canceled (44%). A substantial portion were no show (34%), followed by repeat ED encounters (12%), failed to make any follow-up appointment (5.8%), or completed after the time period ordered by the ED physician (4.4%). These outcomes are summarized in Table 3. Over the 31 months of this study, 71% of the individual patients had a single emergent encounter related to PDR. The percent of patients with 2 encounters (23%), 3 encounters (4.2%), and 4 encounters (1.6%) showed a progressive decrease. There was a single patient who had 6 emergent encounters for PDR during the study period.

Table 3.

Follow-up Outcomes after Emergent Visits for Proliferative Diabetic Retinopathy

	Total	Follow-up Non-compliant	Follow-up Compliant
Overall (n, %)	590 (100%)	225 (38.1%)	365 (61.9%)
Individual Outcome (n, %)
Completed	375 (64%)	10 (2.7%)	365 (97.3%)
Canceled	100 (17%)	100 (100%)
No Show	76 (13%)	76 (100%)
ED	26 (4%)	26 (100%)
Not Scheduled	13 (2%)	13 (100%)

Demographic Variables

There was a statistically significant association between follow-up compliance and marital status (Figure 1). Married patients had increased odds of compliance compared to single patients with an OR of 1.507 (95% confidence interval (CI) = (1.024, 2.218), p = 0.04). None of the other demographic characteristics, including race, primary language, ethnicity, gender, age, and median household income, had a significant OR, which is shown in Figure 1.

Figure 1.

Odds Ratios for Demographic Variables and Follow-up Compliance

Median Household Income refers to the median household income for the patient’s home zip code. The OR is for $5,000 increments of increasing income and a ratio of greater than 1 would indicate a higher likelihood of follow-up compliance. The OR for a patient’s age is calculated for a 10-year increment of increasing age. For marriage status, single was the comparator for ORs. For Primary Language, English was the comparator for ORs. For race, Black/African American was the comparator for ORs. For Ethnicity, non-Hispanic was the comparator for the OR. For gender, male was the comparator for the OR. Statistically significant OR: *p < 0.05.

Statistically significant OR: *p < 0.05.

Visit Variables

There was a statistically significant association between follow-up compliance and specific payers. Patients with Medicaid had reduced odds of compliance compared to all other payers with an OR of 0.543 (95% CI = (0.360, 0.818), p = 0.004). Patients with PHT insurance had increased odds of follow-up compared to all other payers with an OR of 2.749 (95% CI = (1.671, 4.524), p < 0.0001). The other insurance payers (commercial, Medicare, self-pay) did not have significant ORs. These results are represented in Figure 2. Though worsening vision in either eye showed a trend toward increased compliance, these ORs did not reach statistical significance (Figure 2). Initiating treatment with PRP or IVI did not influence the odds of follow-up, nor did the distance between the patient’s home and the clinic. Patients with longer emergent encounters had reduced odds of completing a follow-up appointment, with an OR of 0.948 for each half hour increase in initial encounter time (95% CI = (0.906, 0.992), p = 0.001). Patients with longer intervals to follow-up visits had reduced odds of completion with an OR of 0.941 for a one week increase in interval to follow-up (95% CI = (0.925, 0.956), p < 0.0001) (Figure 2.).

Figure 2.

Odds Ratios for Encounter Variables and Follow-up Compliance

Time to follow-up refers to the number of days after the ED encounter until the next encounter. The OR is for increasing by an increment of 7 days, and a ratio of less than 1 indicates a lower likelihood of follow-up compliance. ED visit length refers to the duration of the ED encounter from check-in to discharge in minutes. The OR is calculated for increasing by an increment of 30 minutes. Worse eye VA and better eye VA refer to the visual acuity of the worse seeing eye at the time of the ED encounter and the better seeing eye at the time of the ED encounter respectively. Both of the ORs are calculated for increasing by an increment of 1 logMAR unit. Distance to ED refers to the distance from the patient’s home zip code to the BPEI ED. The OR is for calculated for an increment of 5 miles. For any treatment emergently, which included PRP to either/both eyes or IVI of anti-VEGF to either/both eyes, no treatment was the comparator for the OR. For insurance payers, the average of all other payers was the comparator for ORs.

Statistically significant ORs: **p < 0.01; ***p < 0.001

DISCUSSION

This study identified a large cohort of patients presenting emergently with active PDR with the goals of determining the rate of follow-up and of identifying factors associated with follow-up compliance. The data show that the follow-up rate for this subset of patients is suboptimal across a wide range of variables. Despite the impetus to access emergency medical care, nearly 40% of PDR patients did not follow up. Though limited data exits, previous studies suggest that patients with ophthalmic complaints are very likely to follow-up after ED encounters, with follow-up rates reported as high as 98%.¹⁶ These rates are much higher than established rates for general ED patients (59%) and rates for other subspecialty referrals such as orthopedics (73.9%).^17,18 Importantly, the low rate found in this study is despite the ability to schedule a follow-up appointment at the time of discharge from the ED, which other studies have proved to be effective at increasing follow-up compliance.^18,19 An immediate follow-up non-compliance rate of 38.1% suggests that PDR patients are a subset of ophthalmic patients at remarkably high risk for being lost to follow-up. Not surprisingly, there is significant overlap between the factors associated with poor follow-up after general ED visits and the factors associated with developing PDR: lower socioeconomic status, less education, lack of insurance, and poor health literacy.^20,21 This scenario, where the same barriers contribute to the development of the disease while also encumbering access to necessary treatment, creates an urgent need for intervention.

The low rate of follow-up is particularly troubling given the severe vision loss observed. Patients in this study had a median VA in one eye reduced to 1.52 logMAR (20/800 Snellen equivalent) and the median VA in the better-seeing eye reduced to 0.48 logMAR (20/60 Snellen equivalent). Such vision loss highlights the chronicity and severity of diabetes among the cohort. This disease burden affects segments of the general population unevenly. Though the catchment area for BPEI is larger than MDC, 86% of emergent PDR visits were for residents of this county. Compared to MDC overall, these patients were significantly more likely to be male, non-Hispanic, Black or African American, older, and have lower income. Previous studies have shown that income is linked to having received eye care in general and that race is significantly associated with poor glycemic control and the progression to PDR.^20,22,23,24 Others have found that vulnerable populations affected by socioeconomic and racial inequalities are accessing care through EDs at increasing rates.²⁵ While sex, ethnicity, race, age, and household income did not have significant associations with follow-up compliance in this study, these variables appear to be important determinants of the antecedent need for emergency care. Rather than exerting a differential effect, the same characteristics that influence the development of PDR continue to hamper follow-up care, largely independent of other demographic characteristics. Notably, marital status was associated with increased odds of follow-up. In the setting of severe vision loss, such support is critical for navigating the health system – a finding consistent with other research demonstrating the positive health impacts of marriage.²⁶

Two insurance payers were significantly associated with follow-up compliance. Across all variables, the PHT insurance was associated with the highest OR while Medicaid was associated with the lowest OR. Both of these associations were statistically significant with p values of <0.001 and 0.004 respectively. Such divergent outcomes were observed despite many similarities between the PHT and Medicaid populations. They are both government-sponsored insurance programs for patients with demonstrable financial need. Of note, PHT is available regardless of immigration status, but providers are limited to clinics associated with the Jackson Memorial Health system. One possible rationale for the different outcomes is that PHT eliminates some of the uncertainty that may lead to patients avoiding health care. Medications, procedures, and surgeries are all covered. The PHT medical system is limited, but the structure keeps patients within a relatively compact network of providers. Conversely, the fragmented network of physicians that accept Medicaid creates an additional barrier for that at-risk population, impeding follow up. Though a retrospective study is susceptible to selection bias, it is concerning to find that a blinding disease is inadequately cared for in a specific insurance group. Another study, though focused on a different pathology, showed that Medicaid patients are less likely to have undergone appropriate testing for open angle glaucoma.²⁷ The necessary conclusion from these data is that the nature of insurance coverage, apart from its presence or absence or the specific population it covers, is correlated with health outcomes. In order to better manage the growing problem of diabetic retinopathy, ensuring that insurance coverage enables, rather than hinders, access to care should be fundamental.

This study also suggests that particular aspects of an emergency encounter are significantly associated with follow-up compliance. As the length of the ED encounter increased, the OR for successful follow-up decreased. Similarly, a longer interval to the next scheduled appointment was also associated with follow-up non-compliance. Both of these effects were small but significant, and suggest that systemic inefficiencies contribute to the barriers facing these patients. Interestingly, treatment with PRP or IVI was not significantly associated with follow-up. No significant effect was seen if these treatments were considered separately either, though the number of patients in each group was notably small (see Table 2). This is in contrast to a larger study that showed a significantly higher rate of loss to follow-up after PRP treatment compared to IVI (28.0% to 22.1%).²⁸ However, that study defined loss to follow-up as an interval exceeding 12 months between clinic visits. Both studies found high rates of unsuccessful follow up, suggesting treatment during an emergent encounter may be the only opportunity to treat a large subset of PDR patients. Accordingly, treating these patients with PRP – a long acting modality – may be preferable to a single anti-VEGF injection. Though data suggest some advantages of serial IVIs of VEGF inhibitors over PRP, even those studies cautioned that “patient-specific factors, including anticipated visit compliance” are important considerations in choosing treatment.²⁹ The data presented here emphasize how carefully physicians should consider barriers to care when evaluating and treating emergent PDR patients.

The current study has several limitations. As a retrospective chart review, the data are limited by the quality of information documented at the time of presentation. Furthermore, it was exploratory research of many different variables simultaneously, increasing the chance of making a type I error. Though the effect of PHT insurance and time to follow-up are still significant with a “Bonferroni correction” for the 23 tests, we did not judge this correction to be necessary given the type of research.³⁰ The primary outcome in this study was based on a single encounter, which does not entirely reflect the level of sustained care necessary to treat PDR. This endpoint also ignores those patients who are able to continue care after missing the immediate follow-up appointment. However, amongst a population with severe vision loss from a chronic disease, any lapses of care are important, particularly immediately after being impelled to seek help. Finally, this study looked at patients seen in an ophthalmic ED, which is analogous but not identical to other walk in ophthalmology clinics and is certainly different than a general medical ED setting. These differences may make the results of this study less generalizable.

Overall, this study demonstrates that patients seen in an emergency setting for active PDR have a low rate of follow-up compliance despite profound vision loss. Though PDR affects some groups more than others, once affected by the disease patient demographics largely do not have a significant impact on follow-up. Married patients and those with PHT insurance are more likely to follow-up, whereas patients with Medicaid are less likely. Scheduling prompt follow-up after a thorough, efficient initial encounter increases the likelihood of compliance. Given the risk demonstrated in this study, physicians should carefully consider follow-up compliance when treating and counseling emergent patients with PDR.