Loss to follow-up and increase in cup-to-disc ratio among glaucoma patients in the IRIS^® Registry (Intelligent Research in Sight)

Abstract

Purpose:

To assess the association between loss to follow-up (LTFU) and risk of increase in cup-to-disc ratio (CDR) among a national registry cohort of patients with primary open-angle glaucoma (POAG).

Design:

Retrospective longitudinal cohort study.

Participants:

Patients with a POAG diagnosis in 2014 and had CDR documented in the IRIS Registry (Intelligent Research in Sight) both in 2014 and 2019.

Methods:

LTFU was defined as a calendar year or more without an encounter. Log-Poisson regression models were used to estimate the relative risk (RR) and 95% confidence intervals (CIs) for increase in CDR. A sensitivity analysis was also conducted to address ceiling effects for patients with baseline CDR ≥0.8.

Main Outcome Measures:

Increase by ≥0.2 in CDR from 2014 to 2019.

Results:

Among 208,517 patients, 6.9% had an increase by ≥0.2 in CDR from 2014 to 2019. While most patients (81.6%) maintained follow up every year, 16.1% were LTFU for 1–2 years, and 2.3% were LTFU for 3–4 years. A lapse of 3–4 years was associated with 15% higher risk of increase in CDR (adjusted RR [aRR]=1.15, 95% CI: 1.00–1.32) compared to no lapse in care after accounting for age, sex, race, insurance, smoking status, glaucoma severity, baseline intraocular pressure (IOP), and baseline CDR. For patients with CDR≥0.8 at baseline, risk of increase in CDR by ≥0.05 was also associated with LTFU (aRR=1.34, 95% CI: 1.11–1.60; lapse of 3–4 years compared to no lapse).

Conclusions:

LTFU is an independent risk factor for increase in CDR among patients with POAG.

Introduction

Primary open-angle glaucoma (POAG) is characterized by loss of retinal ganglion cells (RGCs) and thinning of the retinal nerve fiber layer (RNFL) that lead to structural changes to the optic nerve head and visual field loss.¹ While glaucoma progression typically occurs gradually, the disease can advance to threaten central visual function.² In fact, 3.61 million people worldwide were estimated to be blind from glaucoma in 2020.³

Unfortunately, many glaucoma patients do not maintain consistent follow-up for monitoring and treatment of their disease.^4–9 Half of POAG patients in a national cohort lapsed in care by a year or more between 2014 and 2019.^10,11 Moreover, single center work has suggested that up to two-thirds of glaucoma patients who return after a lapse in care do so with disease progression or a late complication, suggesting that LTFU leads to worsening disease.⁵

One measure of worsening glaucoma progression is direct assessment of the optic nerve head over time. Assessment of the optic nerve includes an estimate of the ratio between the neuroretinal rim (disc) and the central depression (cup), summarized as a cup-to-disc ratio (CDR), which increases as glaucoma progresses. Larger increases in CDR correspond with greater degree of glaucomatous progression than smaller increases.^1,12 Accordingly, changes in CDR and other optic disc characteristics have defined glaucoma progression endpoints in clinical trials.¹³

The American Academy of Ophthalmology (Academy) IRIS^® Registry (Intelligent Research in Sight), the largest ophthalmic clinical registry in the United States,^14,15 provides a robust platform for analyzing longitudinal measurements of CDR. Additionally, the IRIS Registry permits assessment of follow-up patterns over time and has been leveraged to measure LTFU in chronic eye diseases,^16,17 including POAG.^10,11

The purpose of this study was to examine if LTFU is an independent risk factor for increase in CDR among a national cohort of POAG patients in the IRIS Registry.

Methods

The IRIS Registry is a centralized data repository and reporting tool that can be used for research purposes. This does not constitute human subject research because data in the IRIS Registry is de-identified and the investigator does not have access to study identifiers. Therefore, informed consent is not required. Accordingly, the University of Pittsburgh Institutional Review Board determined this study to be exempt from review. We adhered to the tenets of the Declaration of Helsinki.

Study Population

Our study cohort was derived from the IRIS Registry, the largest ophthalmic clinical database in the United States.^14,15 Data from about 12,000 ophthalmologists are represented in the database, which include demographic information, procedure and International Classification of Diseases (ICD) diagnosis codes, and clinical information in the form of visual acuity (VA), intraocular pressure (IOP), and CDR.^14,15 Our cohort from the IRIS Registry includes patients with a POAG diagnosis code in the year 2014, the first full year that data are available in the repository with as much as one year’s past history, and the cohort was examined though 2019. For this study, we included only adult patients and excluded patients who had ICD-9 or ICD-10 codes for angle-closure glaucoma, secondary glaucoma, other optic neuropathy, or glaucoma suspect in a lookback period from 2000–2014. The ICD-9 and ICD-10 codes used to define our cohort have been described previously.¹⁰ For this study on CDR outcome, we included only patients with at least one documented CDR in 2014 and 2019 (Figure 1).

Figure 1:

Flowchart of study population selection from the IRIS Registry

Outcome Definition

The primary outcome was increase in CDR by ≥0.2 from 2014 to 2019. We selected a change of ≥0.2 to define a significant increase in CDR because this threshold has previously been defined to mark true progression rather than observer variability.^18,19 We extracted CDR from the right eye unless only data for the left eye were available, in which case left eye data were used. All documented CDR values in 2014 and 2019 were averaged for a mean annual CDR per patient. Change in CDR was defined as the difference between mean CDR from 2019 and mean CDR from 2014.

For patients with a baseline CDR≥0.8 in 2014, we conducted a separate sensitivity analysis because a ceiling effect would preclude this population from reaching an increase of ≥0.2. For this CDR≥0.8 subcohort, the outcome for increase in CDR was defined as an increase by 0.05 or more, as even small increases in CDR can have substantial functional implications for patients with a cupped disc.^1,20 In fact, the nonlinear relationship between increasing CDR and loss of RGCs makes changes particularly consequential for those with a high baseline CDR. Specifically, a 0.2 change in CDR from 0.5 to 0.7 corresponds to a loss of about 130,000 RGCs, while a 0.2 change from 0.7 to 0.9 corresponds to a much larger loss of 350,000 RGCs.²⁰ Assessment of CDR at the level of 0.05 increments has been associated with increased sensitivity to change and reduced test-retest variability compared to 0.1 increments.²¹

Baseline Variables

Baseline characteristics were obtained from the first visit in 2014. Patient-level data included age, sex, race/ethnicity as reported by practices in the electronic health record (EHR), insurance type, geographic region of patient residence, and smoking status. Clinical history included glaucoma severity as documented by ICD-9 or ICD-10 code and history of glaucoma procedures as defined by current procedural terminology (CPT) codes.¹⁰ Eye-level variables included VA and IOP at the time of first visit. VA measurements were expressed in the logarithm of the minimum angle of resolution (logMAR) and categorized into three groups: under 0.3 (equivalent to better than 20/40 on the Snellen scale, classified as normal), 0.3 to 1.0 (ranging from 20/40 to 20/200, classified as visual impairment), and 1.0 or greater (equivalent to 20/200 or worse, classified as legal blindness).

Loss to Follow-Up

We defined LTFU as the absence of an encounter for at least one calendar year over the study period. We chose a 1-year threshold to define LTFU based on Academy Preferred Practice Pattern^® Guidelines, which recommend follow up for POAG at least annually.²² To avoid dependence on billing codes, we defined an encounter as any documentation of VA, IOP, or CDR on a given date, as we have defined previously.^10,11 Therefore, documentation of at least one exam element by a physician to the IRIS Registry was considered an encounter for follow-up. LTFU was defined as a calendar year or more without at least one documented VA, IOP, or CDR. We subcategorized LTFU by the duration of the lapse in care, as measured by the number of consecutive calendar years without an encounter between the 2014 and 2019 timepoints. For patients with multiple episodes of LTFU, the longest lapse was used for analysis.

Statistical Analysis

Continuous data were summarized as mean ± standard deviation (SD) and categorical data were summarized with frequencies and percentages. Univariable and multivariable log-Poisson regression models were used to estimate the relative risk (RR) and 95% confidence interval (CIs) of rapid increase in CDR associated with baseline demographics and clinical characteristics, with particular attention to intervals of LTFU. Confounders adjusted in the multivariable model were selected from significant variables from the unadjusted model and from variables that have previously been established as risk factors for glaucoma progression. An interaction term between each baseline confounder and LTFU was included in the adjusted model to assess potential effect modification using the “glht” function in R (version 4.4.1; R Foundation for Statistical Computing, Vienna). All analyses were conducted using R, with two-sided tests, and a P-value <0.05 was considered statistically significant.

Results

Study Sample

A total of 804,805 adult patients with a POAG diagnosis code in the IRIS Registry in the year 2014 were extracted for analysis, of whom 607,143 were left after excluding patients with diagnosis codes for angle closure or nonglaucomatous optic neuropathy, and 208,517 patients remained for analysis after requiring at least one documented CDR in both 2014 and 2019 (Figure 1).

The study population had a mean (SD) age of 70.3 (11.4) years, had a slight female predominance (58.1%), identified mostly as White race (68.3%), and most commonly carried Medicare fee-for-service insurance (67.5%) (Table 1). The mean ± SD baseline CDR was 0.57± 0.20 among the entire cohort, 0.54 ± 0.19 for mild-stage POAG, 0.63 ± 0.18 for moderate-stage, 0.76 ± 0.18 for severe-stage, and 0.57± 0.20 for unspecified stage.

Table 1.

Unadjusted associations between patient characteristics and increase in cupto-disc ratio

		CDR increase ≥0.2 (n=14,359/208,517; 6.9%)
Variables	n (%)	Relative Risk (95% CI)	p value
Loss to follow-up:
Longest lapse between encounters			0.02
0 (had an encounter every year)	170190 (81.6%)	Reference
1 year	26193 (12.6%)	0.96 (0.91, 1.01)
2 years	7211 (3.5%)	1.06 (0.97, 1.16)
3 years	2965 (1.4%)	1.18 (1.04, 1.34)
4 years	1958 (0.9%)	0.93 (0.77, 1.10)
Longest lapse between encounters (binned)			0.26
0 (had an encounter every year)	170190 (81.6%)	Reference
1 −2 years	33404 (16.0%)	0.98 (0.94, 1.03)
3–4 years	4923 (2.4%)	1.08 (0.97, 1.20)
Age (years)			<0.001
<60	33124 (15.9%)	Reference
60–<70	58536 (28.1%)	1.16 (1.1, 1.23)
70–<80	71154 (34.1%)	1.39 (1.32, 1.47)
80+	45703 (21.9%)	1.69 (1.60, 1.79)
Sex			0.008
Female	121192 (58.1%)	Reference
Male	86767 (41.6%)	1.05 (1.01, 1.08)
Missing	558 (0.3%)
Race			<0.001
Asian	6123 (2.9%)	Reference
Black or African American	28076 (13.5%)	1.09 (0.98, 1.22)
White	142457 (68.3%)	1.20 (1.08, 1.34)
Hispanic	15204 (7.3%)	1.15 (1.02, 1.29)
Native American and Alaska Native	716 (0.3%)	0.78 (0.53, 1.09)
Native Hawaiian and Other Pacific Islander	285 (0.1%)	0.77 (0.42, 1.28)
Missing	15656 (7.5%)
Insurance			<0.001
Private	34245 (16.4%)	Reference
Govt	2692 (1.3%)	1.24 (1.06, 1.45)
Medicaid	3458 (1.7%)	1.19 (1.03, 1.36)
Medicare FFS	140700 (67.5%)	1.40 (1.33, 1.47)
Medicare Managed	22137 (10.6%)	1.17 (1.09, 1.25)
Military	902 (0.4%)	1.17 (0.89, 1.51)
Unknown/Missing	4383 (2.1%)
Patient region			<0.001
Midwest	35824 (17.2%)	1.03 (0.97, 1.10)
Northeast	24940 (12.0%)	1.15 (1.09, 1.21)
South	67314 (32.3%)	1.09 (1.03, 1.16)
West	30228 (14.5%)	Reference
Unknown/Missing	50211 (24.1%)
Smoking status			<0.001
Never	111989 (53.7%)	Reference
Former	70732 (33.9%)	1.20 (1.16, 1.24)
Active	23348 (11.2%)	1.17 (1.11, 1.23)
Unknown/Missing	2448 (1.2%)
Glaucoma severity			0.04
Mild	12336 (5.9%)	Reference
Moderate	8423 (4.0%)	1.09 (0.99, 1.21)
Severe	4896 (2.3%)	0.97 (0.91, 1.04)
Unspecified	182862 (87.7%)	0.95 (0.84, 1.08)
Visual acuity, logMAR			<0.001
<0.3	150376 (72.1%)	Reference
0.3–<1.0	43915 (21.1%)	1.32 (1.27, 1.37)
≥1.0	8066 (3.9%)	1.59 (1.48, 1.71)
Missing	6160 (3.0%)
IOP, mmHg ≥19			<0.001
Yes	43430 (20.8%)	1.25 (1.20, 1.30)
No	126024 (60.4%)	Reference
Missing	39063 (18.7%)
CDR ≥0.70			<0.001
Yes	78870 (37.8%)	0.24 (0.23, 0.25)
No	129647 (62.2%)	Reference
History of any glaucoma surgery or laser			0.002
Yes	8555 (4.1%)	1.13 (1.05, 1.22)
No	199962 (95.9%)	Reference
History of laser trabeculoplasty			0.05
Yes	6246 (3.0%)	1.10 (1.00, 1.20)
No	202271 (97.0%)	Reference
History of glaucoma surgery			0.002
Yes	2594 (1.2%)	1.24 (1.08, 1.41)
No	205923 (98.8%)	Reference

CDR, cup-to-disc ratio; CI, confidence interval; FFS, fee for service; Govt, government (including the Indian Health Service and state-specific non-Medicaid plans); IOP, intraocular pressure; logMAR, logarithm of the minimum angle of resolution; mmHg, millimeters mercury.

The Northeast region comprises ME, VT, NH, MA, RI, CT. NY, NJ, PA; South: MD, DE, VA, WV, NC, SC, GA, FL, AL, TN, KY, MS, LA, TX, AR, OK; Midwest: MI, OH, IN, IL, WI, MN, IA, MO, ND, SD, NE, KS; West: MT, WY, CO, NM, ID, UT, AZ, NV, CA, OR, WA.

The mean CDR increased from 0.57 ± 0.20 in 2014 to 0.60 ± 0.21 in 2019 (P<0.01) (Figure 2). Increase in CDR by ≥0.2 occurred in 6.9% of the study cohort (14,359/208,517).

Figure 2:

Histogram of cup-to-disc ratio distribution in 2014 compared to 2019

LTFU and Risk of Increase in CDR

While most patients included in this study maintained an encounter at least once every calendar year, 18.4% were LTFU. The most common lapse was one calendar year (12.6%), followed by two (3.5%), three (1.4%), and four (0.9%) consecutive years without an encounter (Table 1).

Patients with a lapse of 3 years LTFU had an 18% higher risk of increase in CDR (RR=1.18, 95% CI: 1.04–1.34) compared to those without lapse. A similar trend was noted when lapses of 3 and 4 years were binned together in the unadjusted analysis (RR=1.08, 95% CI: 0.97–1.20) (Table 1). The association between LTFU and increase in CDR persisted in an adjusted model that accounted for age, sex, race and ethnicity, insurance status, smoking history, glaucoma severity, baseline IOP, and baseline CDR (adjusted RR [aRR]=1.15, 95% CI: 1.00–1.32 for a lapse of 3–4 years compared to no lapse) (Table 2).

Table 2.

Adjusted associations between patient characteristics and increase in cup-to-disc ratio

	CDR increase ≥0.2
Baseline Variables	Adjusted RR (95% CI)
Loss to follow-up:
Longest lapse between encounters
0 (had an encounter every year)	Reference
1–2 years	1.01 (0.96, 1.07)
3–4 years	1.15 (1.00, 1.32)
Age (years)
<60	Reference
60–<70	1.01 (0.93, 1.09)
70–<80	1.24 (1.15, 1.34)
80+	1.61 (1.49, 1.74)
Sex
Male	1.08 (1.04, 1.12)
Female	Reference
Race
Asian	Reference
Black or African American	1.06 (0.94, 1.21)
White	0.98 (0.87, 1.10)
Hispanic	1.07 (0.93, 1.22)
Native American and Alaska Native	0.65 (0.43, 0.95)
Native Hawaiian and Other Pacific Islander	0.86 (0.47, 1.44)
Insurance
Private	Reference
Govt	1.31 (1.09, 1.57)
Medicaid	1.18 (1.00, 1.39)
Medicare FFS	1.28 (1.19, 1.37)
Medicare Managed	1.07 (0.98, 1.17)
Military	1.19 (0.86, 1.61)
Smoking status
Never	Reference
Former	1.12 (1.08, 1.17)
Active	1.19 (1.12, 1.27)
Glaucoma severity
Mild	Reference
Moderate	1.27 (1.13, 1.43)
Severe	1.05 (0.97, 1.13)
Unspecified	1.86 (1.61, 2.14)
IOP≥19 mmHg
Yes	1.17 (1.13, 1.22)
No	Reference
CDR ≥0.70
Yes	0.23 (0.22, 0.25)
No	Reference

CDR, cup-to-disc ratio; CI, confidence interval; FFS, fee for service; Govt, government (including the Indian Health Service and state-specific non-Medicaid plans); IOP, intraocular pressure; mmHg, millimeters mercury; RR, relative risk.

Baseline characteristics associated with increase in CDR included older age (aRR=1.61, 95% CI: 1.49–1.74 for age 80+ years compared to <60 years), government (rather than private) insurance (aRR=1.31, 95% CI: 1.09–1.57), and active smoking status (aRR=1.19, 95% CI: 1.12–1.27). Clinical risk factors included moderate glaucoma severity (aRR=1.27, 95% CI: 1.13–1.43) or unspecified disease stage (aRR=1.86, 95% CI: 1.61–2.14) compared to mild stage and IOP≥19mmHg (aRR=1.17, 95% CI: 1.13–1.22) (Table 2). In an analysis of effect modification of LTFU by baseline variables, no significant interaction was noted (data not shown).

For patients with a CDR≥0.8 or more at baseline, the sensitivity analysis identified LTFU as an independent risk factor for the CDR outcome of an increase ≥0.05. In an adjusted model that accounted for age, sex, race and ethnicity, insurance status, smoking history, glaucoma severity, baseline IOP, and baseline CDR, LTFU was associated with a 34% higher risk of increase in CDR (aRR=1.34, 95% CI: 1.11–1.60 for a lapse of 3–4 years compared to no lapse). In this sensitivity analysis, LTFU by 3–4 years was second only to age 80 years+ as the greatest risk factor for increase in CDR (Table 3).

Table 3.

Adjusted associations between baseline variables and increase in cup-to-disc ratio among patients with a cup-to-disc ratio of 0.8 or more at baseline

Baseline Variables	Among CDR≥0.8 at baseline Change ≥0.05 Relative Risk (95% CI) (n=43,662)
Longest lapse between encounters
0 (had an encounter every year)	Reference
1 −2 years	1.04 (0.95, 1.13)
3–4 years	1.34 (1.11, 1.60)
Age (years)
<60	Reference
60–<70	1.07 (0.95, 1.20)
70–<80	1.24 (1.11, 1.39)
80+	1.41 (1.26, 1.58)
Race
Asian	Reference
Black or African American	1.07 (0.92, 1.25)
Caucasian	1.00 (0.87, 1.16)
Hispanic	1.03 (0.87, 1.22)
Native American and Alaska Native	0.93 (0.58, 1.43)
Native Hawaiian and Other Pacific Islander	0.55 (0.14, 1.45)
Glaucoma severity
Mild	Reference
Moderate	1.22 (1.02, 1.46)
Severe	1.10 (0.92, 1.31)
Unspecified	1.01 (0.87, 1.18)
Visual acuity, logMAR ≥0.3
No	Reference
Yes	1.07 (1.01, 1.14)
Insurance
Govt	1.15 (0.86,1.50)
Medicaid	1.20 (0.95,1.50)
Medicare FFS	1.12 (1.01,1.23)
Medicare Managed	1.20 (1.06,1.35)
Military	0.98 (0.57,1.55)
Private	Reference

CDR, cup-to-disc ratio; CI, confidence interval; FFS, fee for service; Govt, government (including the Indian Health Service and state-specific non-Medicaid plans); logMAR, logarithm of the minimum angle of resolution.

Discussion

In our retrospective, cohort study of POAG patients in the IRIS Registry, we examined the association between LTFU and increase in CDR by ≥0.2 from 2014 to 2019. Over this study period, 6.9% of the cohort experienced the outcome of an increase in CDR. The risk of increase in CDR was significantly associated with LTFU, with a lapse in care of 3–4 years conferring a 15% greater risk of increase in CDR compared to having no lapse in care in an adjusted model that accounted for other risk factors. In a sensitivity analysis for patients with CDR≥0.8 at baseline that defined an increase in CDR as a change of ≥0.05, LTFU for 3–4 years was associated with a 34% higher risk of CDR progression compared to no lapse in care.

While interpretation of CDR is subjective and faced with inter- and intra-observer variability,^18,23–25 CDR can represent a clinically meaningful metric in assessing the severity and stability of POAG.^13,26 For instance, CDR estimates can differentiate eyes with severe visual field loss from those with more mild or moderate disease,²⁷ and CDR correlates with number of remaining RGCs, the RNFL thickness, and mean deviation on perimetry.^20,28

The risk factors we identified for increase in CDR align with established risk factors for glaucoma progression as measured by perimetry or optical coherence tomography (OCT). For example, among the strongest demographic factors for increase in CDR was advancing age, which has been strongly associated with rapid visual field deterioration.^29–34 Similarly, our identification of elevated IOP at baseline as a risk factor for increase in CDR is consistent with the relationship of higher baseline IOP on more rapid visual field progression.^30–33 Even the relatively smaller effect that we identified between smoking status and CDR progression is consistent with literature on active smoking status and risk of RNFL thinning.³⁵

In addition to assessing known risk factors for glaucoma progression, we identified LTFU as an independent risk factor for increase in CDR. The prevalence of LTFU among POAG patients in the IRIS Registry has been estimated at 50% over 6 years,¹⁰ which aligns with estimates of LTFU from single centers.^5–9 Outcomes of patients with LTFU who eventually return to clinic can provide insight into the clinical consequences of lapses in care. In a single-center study on LTFU over 10 years, a third of patients became LTFU, of whom a sixth returned after a lapse in care of greater than one year.⁵ These patients who returned after a period of LTFU appeared to have poor outcomes, with two-thirds requiring treatment escalation upon their return visit.⁵ Another single-center study of glaucoma patients identified noncompliance with treatment or follow up as a risk factor associated with almost two-fold higher odds of developing blindness in one eye.³⁶ Similarly, patients with POAG in the IRIS Registry were twice as likely to go blind if they lapsed 3–4 years without an encounter compared to patients who maintained regular follow-up.³⁷ In the present study, we leveraged the national cohort of POAG patients in the IRIS Registry to demonstrate that LTFU is independently associated with increased risk of clinical progression of glaucomatous disease, as measured by increase in CDR over a 6-year period.

The associations between baseline characteristics and increase in CDR remained consistent in a sensitivity analysis for the subgroup of patients with CDR≥0.8 at baseline, for whom even a small incremental increase in disc cupping could represent significant loss of RGCs and visual function.^1,20 In fact, LTFU for 3–4 years was second only to age 80 years or more as the strongest risk factor for increase in CDR, highlighting the particular susceptibility to lapses in care among patients with more severe cupping. The consequences of LTFU among a group that represents the most severe disease is concerning because patients with greater disease severity are at highest risk of LTFU.¹⁰ Thus, patients with severe-stage glaucoma are not only at the highest risk of becoming LTFU, but they are also must susceptible to the consequences of lapses in care.

LTFU is addressable, and interventions can help glaucoma patients to maintain care. Appointment no-show is a risk factor for LTFU and a potential opportunity to engage with patients at risk of discontinuation of care.³⁸ Glaucoma patients with an appointment no-show report barriers to care that include transportation and financial challenges, which are barriers that can be successfully resolved through referral to social work or patient navigator programs.^39–42 More simply, a commonly stated reason for missed appointment is forgetting about the scheduled visit,³⁹ and automated messages sent after a missed visit can double the odds of an ophthalmic patient re-engaging with ophthalmic care after an appointment no-show.⁴³

Finally, recent evidence suggests that re-engagement of glaucoma patients through telehealth outreach after LTFU can double the odds of return to eye care.⁴⁴ Additionally, artificial intelligence-enabled methods, such as natural language processing (NLP), can also be deployed within the EHR to identify ophthalmic patients at risk of LTFU by, for instance, flagging language related to health-related social needs like transportation difficulties.⁴⁵ NLP algorithms can detect social needs in parallel to EHR-embedded screening questionnaires to assess risk of LTFU and, ultimately, to facilitate referrals to social resources in order to help maintain care.⁴⁶

Limitations

Despite the strengths of our longitudinal assessment of a large POAG cohort using real-world data, our study has several limitations. Assessment of CDR is subjective, and we examine only the CDR reported by treating ophthalmologists without the ability to confirm their interpretations by fundus photography or reading center criteria. Similarly, it is unknown whether CDR was measured at the vertical meridian, as is most relevant for glaucoma assessment, and we do not have information about the techniques individual ophthalmologists used to measure CDR (such as the power of the condensing lens at slit lamp biomicroscopy or whether ancillary structural image testing, like fundus photos or OCT, were used to aid CDR assessment). To reduce variability in subjective assessments, we averaged the annual CDR for each patient in the analysis. More detailed examination findings of the optic nerve head, such as notches, hemorrhages, or focal rim thinning, are important for longitudinal assessment for glaucoma progression,²² but these examination details are not available in the IRIS Registry dataset. While ancillary testing reports from perimetry or optic nerve imaging studies would greatly enhance the definition of glaucoma progression, these data elements are not presently available in the IRIS Registry for analysis.

Regarding the measurement of lapses in care, it is not known what follow-up intervals were recommended by the treating ophthalmologists, and it is possible that more mild-stage or low-risk POAG patients were informed to follow-up less often than once per year. These low-risk POAG patients may have also been comanaged with a non-IRIS Registry practice, such as another ophthalmic practice or optometry-only group, and thus potentially falsely labeled as LTFU due to the absence of IRIS Registry data. We aimed to mitigate this risk by excluding glaucoma suspect codes and by conservatively considering any visit in the IRIS Registry over a calendar year as maintaining care, although more severe disease would likely warrant closer follow-up. Along the same lines, the IRIS Registry does not contain data from the majority of academic medical centers, nor does it contain records from practices on paper charts, so patients with visits at these settings would be underrepresented. It is also unknown whether a patient’s POAG was necessarily addressed at a given encounter, but we conservatively labeled any encounter as valid contact with an ophthalmologist for the calendar year. Finally, while LTFU is presumably associated with nonadherence to glaucoma medications, we do not have detailed information about glaucoma medication usage, prescriptions, or refills to assess medication adherence.

In conclusion, we identified demographic and clinical risk factors for rapid increase in CDR among POAG patients in the IRIS Registry. LTFU was found to be independently associated with an increase in CDR, suggesting that there may be measurable clinical consequences of lapses in care among a cohort of patients with POAG even over a 6-year period.

Data Sharing Statement:

Data are proprietary