Factors Associated with Adherence to Screening Guidelines for Diabetic Retinopathy Among Low-Income Metropolitan Patients

Abstract

In this retrospective analysis of patients with diabetes in an academic primary care clinic in St. Louis, attendance at ophthalmic screening appointments was recorded over a two-year observation window. Factors associated with adherence were analyzed by multivariable regression. Among 974 total patients included, only 330 (33.9%) were adherent within a two-year period. Multivariate analyses identified older age, female gender, primary language other than English, and attendance at ancillary diabetes clinic visits as factors associated with improved diabetic retinopathy screening adherence. Factors not associated with adherence included race and insurance status.

Introduction

Diabetic retinopathy is one of the leading causes of preventable new-onset vision loss in adults worldwide.¹ Because the disease is clinically silent until very late stages, it is estimated that up to 90% of vision loss may be prevented by regular screening examinations.² Guidelines issues by the American Academy of Ophthalmology (AAO) and the American Diabetes Association (ADA) recommend annual dilated eye examinations in adult patients with diabetes. Recently, these guidelines were amended such that if an initial retinal exam is normal, and glucose control is good, the interval between dilated retinal exams could be extended to two years.³ Even so, only 50%-60% of patients with diabetes in the U.S. adhere to these recommended guidelines.^4–6 In populations with poor access to resources, which tend to be at greatest risk for vision loss from advanced stages of diabetic retinopathy (Figure 1), rates of eye examinations are even lower.^7–11 Patients cite multiple barriers preventing them from attending eye exams, including transportation difficulties, cost, and long wait times.^12–14 Better quantification and qualification of the prevailing factors associated with poor adherence is essential to implementing strategies to mitigate key barriers to retinal examination among patients with diabetes, and improve their rates of adherence to screening guidelines.

Figure 1.

High-risk Proliferative Diabetic Retinopathy. Although diabetes is a leading cause of impaired vision and blindness, regular screening examinations can prevent up to 90% of vision loss from advanced stages of disease as show in this patient with vitreous hemorrhage from proliferative diabetic retinopathy.

Referral to an eye care specialist for in-person retinal examination remains the standard for diabetic retinopathy screening in the U.S. Such practice patterns are currently in place at the Washington University School of Medicine-Barnes Jewish Hospital healthcare network, which serves as a “safety net” for many medically-vulnerable patients in the St. Louis City and County area. Rates of adherence to eye examinations have historically been low in this population, highlighting a need for improvement in health care delivery. In this retrospective study, we sought to 1) determine the rate of adherence to recommended eye screenings; and 2) identify factors associated with adherence to recommended eye care guidelines in a primary care medicine clinic serving predominantly high-risk, low-income patients with diabetes.

Patients, Methods, and Materials

Subjects

Consecutive patients with type 1 or type 2 diabetes mellitus over the age of 18 who completed a non-emergent appointment at the Primary Care Medicine Clinic (PCMC) at Barnes Jewish Hospital between July 2016 and March 2017 were included in the study. Participants were excluded if they lived outside of a 50-mile radius from the PCMC in order to exclude traveling distance as a factor contributing to adherence. Every patient with diabetes was encouraged to follow up with their previous ophthalmologist if established, or referred to the Washington University Eye Clinic for an annual diabetic eye exam in accordance with the screening recommendations delineated by the AAO and the ADA.

Patient data – including demographic information, insurance status (classified as either private, public [Medicaid or Medicare], or self-pay), pertinent laboratory values, and ancillary diabetes clinic visits (foot exams and diabetes education visits) – were extracted from the electronic medical record system and de-identified for analysis. Each patient record was manually reviewed to document instances of eye care clinic attendance within a two-year window, beginning one-year interval before and ending one-year after the reference visit at the PCMC. Patients who had at least one documented eye care visit within this two-year window were denoted adherent with annual eye screening.

Statistical Analyses

Descriptive statistics (mean, standard deviation and percent) were reported for demographics and laboratory data. Various factors were analyzed for their association with patient adherence using SAS 9.4 software (SAS Institute, Cary, NC). To adjust for confounding effects, variables with univariate logistic p-values <0.05 were entered simultaneously into a multivariable regression model. Factors with an adjusted p-value <0.05 were retained. Odds ratios were used to describe the relationship between factors and adherence.

Results

Patient Demographics

Included in this study were 981 unique adult patients with diabetes who visited PCMC during the specified time period. After excluding seven patients because their primary residence was greater than 50 miles from the PCMC, 974 patients were included in the final analysis. The study population was predominantly female (58.0%), African-American (76.6%), and with an average age of 56 years (Table 1). The majority of patients (70.6%) were covered by public insurance (Medicare or Medicaid).

Table 1.

Patient Demographics with Chi-square Analysis

Demographics	Patients n = 974 (% of total)	Adherence† n = 330 (% adherent)	p-value	OR 95% CI††
Age
Mean	56; SD = 13.8	58.4; SD = 13.6	0.0022§	1.49 (1.12, 1.98)§
Sex
Male	409 (42)	118 (28.9)	0.005	1.48 (1.13, 1.95)
Female	565 (58)	212 (37.5)
Race
White	161 (16.5)	55 (34.2)	0.259	0.95 (0.66, 1.36) 1.46 (0.81, 2.63)
Black or Afr Am	746 (76.6)	246 (33.0)
Other†	65 (6.7)	28 (43.1)
Unknown	2 (0.2)	1 (50.0)
Insurance Type‡
Public	688 (70.6)	248 (36.1)	0.0208	1.73 (1.17, 2.57) 1.44 (0.84, 2.46)
Private	110 (11.3)	35 (31.8)
Self-pay	159 (16.3)	39 (24.5)
Unknown	17 (1.7)
Primary Language
English	904 (92.8)	296 (32.7)	0.0079	0.52 (0.32, 0.84)
Not English	70 (7.2)	34 (48.6)
Ancillary Visits#
No visit	860 (88.3)	271 (31.5)	<0.0001	2.33 (1.57, 3.46)
≥1 visit	114 (11.7)	59 (51.8)

^*Asian, Hispanic, and of Middle Eastern descent

^†Adherence defined as at least one ophthalmology office visit within two-year time period. Percentages of each subgroup within a demographic recorded as adherent

^††CI calculated as odd ratio of adherence to nonadherence within subgroups, excluding the unknown values.

^‡Public insurance included Medicare and Medicaid and Private included all PPO and HMO. N=17 patients excluded because insurance type listed as “miscellaneous” or “commercial.”

^§p-value and OR from chi-square analysis reflects age per decade comparison between adherent and nonadherent groups

^#Ancillary diabetes visits are calculated as the sum of diabetes education visits and foot care visits

Bolded = significant CI.

Predictors of Adherence by Univariate Chi-Square Analysis

Of the 974 patients included, 330 distinct patients were adherent with ophthalmic examination. Characteristics associated with higher adherence in univariate analysis included older age, female sex, public insurance, and attendance to at least one other diabetes care visits (p<0.05; Table 1). The odds of women being adherent were 1.48 the odds of men being adherent at the time of their visit to the primary care clinic. Patients with public insurance had 1.75 times the odds of those without insurance coverage to have a screening diabetes eye exam. However, there was no statistically significant difference between adherence rates in patients with private insurance versus any other insurance subgroup. Patients who attended at least one associated diabetes care visit (foot care visit or diabetes education visit) had about 2.3 times the odds of having attended a screening eye exam as those who did not (OR 2.33 [95% CI, 1.57–3.46]). Unexpectedly, English speakers were about half as likely as non-English speakers to complete recommended retinal screening.

Predictors of Adherence by Multivariable Analysis

Factors including insurance type, age per decade, female gender, attendance at ancillary diabetes visits, and primary language other than English were entered into the multivariable analysis. Insurance type was no longer a significant factor after adjusting for the effects of age (aOR 1.17 [95% CI, 1.06–1.29]), female sex (aOR 1.48 [95% CI, 1.11–1.96]), ancillary diabetes education visits (aOR 2.41 [95% CI, 1.61–3.61]), and primary language (aOR 0.55 [95% CI, 0.33–0.91]). All remaining variables were significantly associated with increased adherence with diabetes eye exams (Table 2). The Hosmer-Lemeshow Goodness-of-Fit test demonstrated good fit in our multivariable model (p=0.75).

Table 2.

Multivariate logistic regression

Predictor	Univariate p-value	Adjusted p-value	p-value; excluding insurance	aOR 95% CI*
Insurance	0.0208	0.0931	--	--
Age per decade	0.0022	0.01	0.002	1.17 (1.06, 1.29)
Sex (Female vs Male)	0.0049	0.0132	0.007	1.48 (1.11, 1.96)
≥1 Ancillary Visit	<0.0001	<0.0001	<0.0001	2.41 (1.61, 3.61)
Primary Language (English vs Not English)	0.0079	0.0088	0.0189	0.55 (0.33, 0.91)

^*aOR = adjusted odds ratios calculated from multivariable regression after excluding insurance type.

Bolded = significant CI.

Race was not noted to be a significant predictor of adherence under univariate analysis and was therefore excluded from multivariate analysis. Similarly, factors such as last serum glycated hemoglobin index (HbA1c) values and last LDL values did not demonstrate a statistically significant difference between the adherent and non-adherent groups (HbA1c: n=313 7.93% ± 1.78 SD vs. n=570 7.95% ± 2.15 SD, respectively; OR 0.996 [95% CI, 0.93–1.07]; LDL: n=279 94.28 mg/dL ± 42.8 SD vs n=468 99.46 mg/dL ± 42.5 SD, respectively; OR 0.997 [95% CI, 0.994–1.00]).

Discussion

Adhering to diabetes eye screening recommendations continues to be a challenging task, particularly in a patient population that is socioeconomically disadvantaged. We sought to highlight this difficulty by quantifying the rate of adherence within a primary care clinic serving low-income urban patients. Previous papers have cited adherence rates in the 50–60% range for patients with diabetes, but populations in these studies tended to contain more affluent and education patients.^{6, 9} Our study demonstrates a far smaller rate of 33.9% adherence, comparable to recent studies based on similar demographic population of predominantly low-income African-American patients with diabetes.^{4, 8, 10}

We chose to utilize a two-year window based on recent evidence that promotes screening diabetes eye examinations every two years, instead of annually.¹⁵ By allowing a two-year window to fulfill the screening diabetes eye examination, we may have overestimated the true adherence rate in our population. When evaluating adherence during the one-year interval following the reference PCP visit, we found that only 18.7% of patients with diabetes fulfilled their recommended screening eye examination. Sub-analysis of this group showed that patients who demonstrated adherence in the past were more than three times as likely to be adherent in the one-year following the reference PCP visit (39.9% vs. 11.5%). These findings highlight the difficulty of changing patterns of nonadherence but suggest that measures to increase adherence to even one eye care visit may produce durable adherent behavior.

Additionally, we found that in our cohort of patients, older age, female gender, attendance at diabetes-related care visits were all associated with greater adherence to diabetic retinopathy screening. In our multivariate logistic regression for adherence, female patients were 1.48 times as likely to be adherent with screening. Likewise, older individuals were more likely to comply with recommended screening eye exams, with our model showing that each additional decade of life increased odds of adherence by 17%.

Among the factors identified in this study linked to increased adherence to annual retinal screening examinations, attendance to ancillary diabetes care visits (diabetes education sessions and foot care visits) was the strongest predictor (OR 2.41). One explanation for these findings is that eye care, education, and foot care visits all have similar barriers preventing patients from attending them. Those able and/or willing to overcome those barriers in one setting are self-selecting to comply with screening guidelines. However, an alternate explanation is that increasing patient self-awareness through educational measures may have the largest impact in determining future adherence to health screening recommendations. In support of the latter view, the beneficial effects of educational programs on multiple aspects of diabetes care, including weight loss, glycemic control, self-monitoring, and adherence to health screening (including retinopathy surveillance) are well-established.^16–20 Furthermore, attendance at diabetes education sessions has previously correlated with increased rates of adherence to screening recommendations.^21–23 One of the health systems with the highest rates of diabetic retinopathy screening adherence – the Veteran’s Health Network, which consistently reports 90% adherence rates for annual dilated retinal examination – utilizes a multipronged educational approach to ensure that patients with diabetes are meeting several metrics of appropriate diabetes care, such as glycated hemoglobin goals, foot examinations, setting blood pressure goals and receiving retinal examinations.²⁴

In other study populations, African-American race, younger age, male sex, and lower household income are associated with lower adherence.^{4, 8, 25, 26} We found no such association with race or income, likely because our population was predominantly poor, African-American (76.6%), and supported by publicly-funded healthcare (70.6%). Our findings are similar to a recent analysis of factors associated with retinopathy screening adherence in a population with comparable demographics.⁸ Unexpectedly, English speakers were nearly two times less likely than non-English speakers to complete recommended retinal screening. Though language barriers and difficulty obtaining interpreter services are potential obstacles to health care access and diabetic retinopathy screening, these factors may have been mitigated in our study population by the provision of these services free of charge through Barnes Jewish Hospital. The interpreter available at PCMC may also serve as a healthcare advocate who can help the patient navigate the referral system.

A major limitation in our retrospective review was the inability to account for screening eye examinations that occurred outside the Barnes Jewish Washington University network. In our study design, patients who underwent eye examination outside of our healthcare network would have been incorrectly categorized as noncompliant, resulting in a falsely-low observed adherence rate. However, based on prior referral patterns within our tertiary care center, shared electronic medical record systems, and identical insurance provider policies, nearly all patients receiving primary care at PCMC also receive ophthalmic care within the Washington University healthcare network, suggesting the number of outside screening exams to be minimal. Furthermore, the utilization of a two-year window to allow for adherence was meant to compensate for any underestimation from non-recorded eye examinations elsewhere.

The results of our study suggest that certain patient characteristics – male gender, younger age, private insurance policy holders or self-pay patients, patients who do not attend other diabetes clinic visits – may correlate with lower adherence with recommended diabetes eye examinations. This knowledge may be helpful in identifying more at-risk patients who may need additional counseling in order to motivate them to adhere to recommended screening eye exams. The results of this analysis highlight the complex network of influences that affect patient adherence as well as the difficulty in overcoming the numerous barriers that patients face in adhering to recommended health care exams. Many of those barriers may be circumvented if screening is conducted during the primary care encounter as a point-of-service test, instead of during a separate clinic visit. These tests could be provided during wait times, or other similar periods of currently underutilized clinic time. In a study of healthcare claims data of Medicaid patients, completion rates of diabetic retinopathy screening examinations were more likely when a non-mydriatic retina camera was available in the primary care setting.²⁷ Automated image analysis coupled to non-mydriatic fundus photography may provide this opportunity, allowing for quick, affordable, and accurate automated screening as a point of care test (Figure 2). Furthermore, point-of-care retinal testing introduces an opportunity for individualized patient education, tailored to severity of retinal disease. Based on our findings, such educational interventions may be most effective in improving retinal surveillance in our healthcare network and in other networks serving low-income patients. Finally, our results suggest that implementation of services to help patients navigate along various facets of medical landscapes, in addition to improving education and point-of-care services, could improve retinopathy screening and perhaps other aspects of comprehensive diabetes care.

Figure 2.

Non-mydriatic Retinal Photography as a Point-of-Care Service. To improve screening rates among patients with diabetes, ophthalmic photography can be performed in physicians’ office and clinics by easily-trained technicians using retina cameras not requiring drops to dilate pupil.

Conclusion

Low screening rates for diabetic retinopathy in this low-income urban population were associated with younger age, male gender, and lack of attendance at other ancillary health screens or diabetes education sessions. Our findings suggest that efforts to improve diabetes education could have the largest impact to improve adherence rate to retinopathy screening in this population, and other patient populations with similar characteristics. In light of recent efforts to incorporate point-of-care services such as non-mydriatic retinal photography in primary care clinics, our results argue that integrating diabetes education and medical advocacy into such systems would be valuable.

Epilogue: COVID-19 Pandemic

Society and medicine now faces an unprecedented and ongoing crisis caused by the severe acute respiratory syndrome-coronavirus-2 (SARS-CoV2), a pathogen that has claimed over 200,000 lives at the time of this writing. The pandemic caused by SARS-CoV2 highlights an imperative need for a paradigm shift in practice patterns across medical specialties. Such a dire need is now at the forefront of discussion because of the immediate health threat posed by a novel pathogen, but the concerns for preventing nosocomial or iatrogenic transmission of diseases are long-standing.

Hospital and health care visits have always posed potential dangers for contagion (e.g. MRSA, hospital-acquired pneumonia, viral infections), especially among vulnerable and indigent populations. Such concerns for preventing transmissible disease add to other needs for paradigm shifts in medicine – including increase access to medical care in populations that typically lack resources, as discussed above and as highlighted by the results of our study. Therefore, changes in medical practice that permit remote provision of care, when appropriate, or those that allow for fewer physical patient encounters could be of great help.

One such shift, embraced by many of our colleagues in light of the SARS-CoV2 pandemic, involves the use of telemedicine. Though telemedicine has inherent limitations in fields such as ophthalmology that rely heavily on physical examination, technical advances in imaging could provide innovative solutions. As an example, the development of facile and reliable non-mydriatic cameras, capable of being operated by unskilled or non-ophthalmic personnel, would allow retinal imaging in patients across a wide variety of settings such as primary-care offices or remote-screening centers. Digital transfer of the images captured by such cameras to reading centers, physicians-on-call, or even artificial intelligence platforms would allow for rapid access to care, and could increase such access to much larger patient populations. Pre-screening in this manner could also allow care providers to identify high-risk patients who require physical care, thereby minimizing the spread of contagious disease among other low-risk individuals.

The SARS-CoV2 pandemic has upended life and society, but perhaps a silver lining to such disaster is that people across sectors are pioneering ways to solve the various new problems we now face. Some of these novel solutions might mitigate other long-standing problems, just as telemedicine might do for improving poor access to care among those suffering from common diseases that were pandemic long-before SARS-CoV2, such as diabetes.