Highlights
-
•
Delayed glaucoma follow-up (DFU) occurred in over half of all glaucoma visits.
-
•
Most patients had stable HVF mean deviation over the follow-up period.
-
•
Higher rate of DFU was generally not associated with visual field progression.
-
•
Longer follow-up intervals than those currently prescribed may be safe for Many glaucoma patients.
Objective
To query current practice for glaucoma follow-up intervals by studying the association between delayed follow-up (DFU) and glaucoma progression.
Design
Retrospective cohort and matched case-control study.
Subjects
Glaucoma patients and suspects at a tertiary care center.
Methods
Electronic health record and Humphrey Visual Field (HVF) data were collected for patients with at ≥5 HVF tests between 2014 and 2023. DFU was defined as exceeding provider-recommended intervals and quantified for each participant as the percentage of visits with DFU, mean days of DFU, and maximum days of DFU. Associations between DFU metrics and HVF progression were analyzed using covariate adjusted multivariable logistic regression models. Subgroup analyses by baseline mean deviation (MD) and a sensitivity analysis of patients with 2 or more HVF tests with case-control matching (age, follow-up duration, number of HVF tests, and baseline MD) were performed.
Main Outcome Measures
HVF progression: MD linear regression slope ≤−0.5 dB/year across all tests.
Results
A total of 1121 eyes from 600 patients were included (mean follow-up: 7.3 years; mean baseline age: 73.7 years; 50.2% female). HVF progression occurred in 19.8% of eyes with baseline MD >−6 dB and 32.8% with baseline MD ≤−6 dB. Any DFU occurred in 53.4% of all visits. In multivariable analyses, percentage of visits with DFU, mean days of DFU, and maximum days of DFU were not associated with HVF progression, including in subgroup analyses. Older age, larger cup to disc ratio, shorter follow-up duration, higher number of HVF tests, and unmet social needs were associated with higher odds of HVF progression (all P <.05). In the matched case-control analysis, a mean DFU of >60 days was significantly associated with higher odds of HVF progression (aOR: 1.95; 95% CI: 1.03-3.68, compared to ≤30 days), and a maximum DFU of >365 days showed a borderline association (aOR: 1.77; 95% CI: 0.97-3.23, compared to ≤120 days) among patients with baseline MD <−6 dB.
Conclusion
These findings suggest that longer follow-up intervals than those currently prescribed may not be associated with VF progression for most glaucoma patients.
INTRODUCTION
G laucoma, the leading cause of irreversible blindness, affects over four million people in the United States. , Regular clinic visits and diagnostic tests are central to disease management, yet this frequent monitoring imposes a significant burden on the healthcare system and worsens disparities in access to care for patients. ,,,
Close monitoring with multiple clinical visits per year is critical for glaucoma patients with unstable or rapidly progressing disease, , but longer follow-up intervals may be appropriate for most glaucoma patients and suspects, who typically experience slow disease progression. ,,,,,,,,,,, For example, the Ocular Hypertension Treatment Study found that only 25% of ocular hypertensive patients developed visual field defects over 20 years. Similarly, retrospective studies report slow rates of visual field mean deviation (MD) progression for the majority of glaucoma patients, with median rates as low as −0.05 dB per year. ,,,,,,,,,
Despite the stability of most glaucoma patients, concerns about delayed detection of progression have led many doctors to perform frequent examinations for every glaucoma patient. This practice may unnecessarily strain the healthcare system and limit resources for patients with unstable conditions. While prior studies link infrequent follow-up to disease progression in unstable patients, ,, there is limited evidence regarding whether conventionally recommended follow-up schedules may be excessive for the broader glaucoma population.
To address this gap, we investigated the relationship between delayed follow-up (DFU) intervals and glaucoma progression using data from a large tertiary academic center. We hypothesize that patients whose follow-up intervals exceed provider recommendations do not have higher odds of disease progression after adjusting for other risk factors. If DFU is not associated with glaucoma progression in some or all groups of patients, it would support longer follow-up intervals for many glaucoma patients, thereby reducing healthcare utilization while maintaining safe and effective care.
METHODS
PATIENTS AND MEDICAL RECORD DATA
This was a retrospective cohort study of patients with open angle glaucoma or open angle glaucoma suspect at the University of California, San Francisco (UCSF). This study was approved by the UCSF Institutional Review Board and adheres to the tenets of the declaration of Helsinki.
Participants were identified using International Classification of Diseases, Tenth Revision (ICD-10) and Systematized Nomenclature of Medicine—Clinical Terms (SNOMED-CT) codes for open angle glaucoma and open angle glaucoma suspect. Electronic health record (EHR; Epic Systems) data for eligible patients seen between January 2014 and December 2023 were extracted. EHR data included appointment dates, demographic characteristics, and clinical characteristics (e.g., intraocular pressure [IOP], central corneal thickness [CCT], and cup-to-disc ratio [CDR]) and a social determinant of health (SDOH) summary metric described below.
If IOP was measured using multiple devices at the same visit, only one value was retained, with preference given in the following order: Goldman applanation tonometry, followed by pneumotonometry, and then iCare tonometry. If multiple IOP readings using the same type of device were recorded (e.g., before and after administration of ocular antihypertensive agents), the first was retained. The SDOH information was taken from the “SDOH Wheel” in the Epic EHR, and a binary summary metric was created and considered present if participants indicated any unmet need in the following areas: food security, financial resource strain, transportation needs, utilities, or housing stability. Missing values in a specific SDOH metric was considered absent of unmet need of this metric.
Diagnostic test data, including Humphrey visual field (HVF, Humphrey Field Analyzer; Carl Zeiss) and spectral domain ocular coherence tomography (OCT; Optovue, Visionix)—the major OCT system used to evaluate glaucoma patients at UCSF—were extracted. Details on the analysis of HVF data are described below.
DEFINING DELAYED FOLLOW-UP
EHR data were used to calculate the time interval between each participant’s successive clinic visits. Because we were interested in identifying patients with true lapses in ophthalmologic care without clinical care, this analysis included every visit with any optometrist or ophthalmologist within UCSF. Patients were classified as having DFU only if they failed to see one of these eye care providers within the interval recommended by their provider at the previous visit.
We used real-world data from the electronic health record (EHR) at UCSF to define the recommended follow-up interval. At UCSF, providers use the EHR to communicate their recommended follow-up interval to scheduling staff, and this recommended interval was extracted. If a provider did not specify a recommended follow-up date in the EHR, the DFU status for that visit was recorded as missing.
To calculate DFU for each visit, we used the formula: Actual Visit Date— (Prior Visit Date + Recommended Follow-Up Interval in Days). Visits within the recommended interval were assigned a DFU value of 0 days. Using all follow-up visits, 3 DFU metrics were derived for each patient: (1) the percentage of visits classified as having any DFU, (2) the maximum number of days of DFU at any visit, and (3) the average number of days of DFU across all visits per participant.
To account for scheduling variability, a 15% grace period was applied when calculating the percentage of visits with DFU. A visit was considered delayed if it occurred more than 115% of the recommended interval after the previous visit. However, the grace period was not applied when calculating the mean or maximum days of DFU. This distinction was made because the binary threshold analysis is more sensitive to minor timing deviations that may not be clinically meaningful, and the grace period helps prevent misclassification of otherwise on-time visits as delayed. In contrast, continuous measures inherently capture the full distribution of delays, making additional adjustment unnecessary and potentially biasing the magnitude of observed differences. Finally, all 3 DFU measures were categorized into approximate tertiles (with exact cutoffs chosen for ease of interpretation) and used as predictor variables for analyses. We evaluated whether the number of EMR-documented cancellation for each type of cancelations (patient cancellation, no-show, provider cancellation, or COVID-related) was associated with glaucoma progression. We reviewed internal and available external clinical records for patients with maximum DFU ≥180 days to identify documented outside care and assess whether receipt of glaucoma care elsewhere during a DFU period could have influenced study outcomes.
VISUAL FIELD TESTING AND DEFINING GLAUCOMA PROGRESSION
Subjects underwent central 10-2, 24-2 and/or 30-2 HVF tests using the Swedish Interactive Threshold Algorithm (SITA, Standard and Fast algorithms) at variable intervals determined by their glaucoma care provider. The mean deviation (MD) from each HVF test was extracted for each participant. To account for a potential learning effect, the baseline HVF test was defined as the one with the better (i.e., less negative) MD value among the first 2 available HVF tests. If the second HVF was considered the baseline test, the first HVF test was excluded from the analysis. Participants with baseline MD >−6 dB were considered to have glaucoma suspect or mild glaucoma, and participants with baseline MD ≤−6 dB were considered to have moderate or severe glaucoma. ,
The primary outcome measure in this study was HVF progression at the eye level. Individual linear regressions were conducted using MD values from all measurements of each eye in the study. An eye was considered to have HVF progression if their MD declined by more than 0.5 dB per year (i.e., MD linear slope ≤−0.5 dB/year) over the study period, regardless of the statistical significance of the slope. This criteria was chosen for its sensitivity in detecting progression within clinical populations. ,, Given the low glaucoma progression rate and finite follow-up interval in this cohort, a trend-based approach was used to maximize sensitivity for detecting subtle visual field changes that may not be captured by event-based criteria, ,,,,,, while also limiting redundancy and multiple testing from highly correlated progression definitions.
STATISTICAL ANALYSIS
All statistical analyses were performed in R (R Foundation for Statistical Computing, Vienna, Austria), and a 2-sided P <.05 was considered statistically significant.
Whole cohort analysis for subjects with 5 or more HVF tests
Participant demographics and clinical characteristics were summarized using mean (SD [SD]) for continuous measures and count (%) for categorical measures. Analyses of the association between each DFU metrics (percentage of DFU, mean days of DFU and maximum days of DFU) and HVF progression (yes/no) were conducted using separate multivariable logistic regression models for each DFU metrics and generalized estimating equations (GEE) were used to account for within-participant inter-eye correlation in the outcome measure. The multivariable models were adjusted for demographics (age, self-reported sex [all study participants reported thier sex as male or female], self-reported race), a binary unmet SDOH need indicator, baseline ocular measures (IOP, CCT, CDR, MD, average RNFL thickness), length of follow-up, and number of HVF tests. The length of follow-up and number of tests—both of which may increase the likelihood of detecting disease progression and may be associated with more unstable or severe glaucoma– were included as covariates because they were not highly correlated (Spearman correlation coefficient <0.5) and could independently confound the association. Adjusted odds ratios (aORs), 95% CIs (95% CI), and P -values from multivariable logistic regression models were reported. To assess potential threshold dependence of the analysis, we repeated the primary analysis using the slope of MD as a continuous outcome, rather than using a binary definition of glaucoma progression.
Subgroup analyses were performed to evaluate whether the associations between DFU and HVF varied based on glaucoma severity, as defined by baseline MD values. In these subgroup analyses, separate multivariable regression models for each glaucoma severity subgroup were performed following the same approach as the primary analysis.
Sensitivity analysis: matched case-control analysis for subjects with 2 or more HVF tests
To validate the findings of the analyses detailed above using the whole cohort with at least 5 HVF tests, we conducted a matched case-control analysis among eyes with at least 2 HVF tests. To maximize the number of matched pairs in the case-control study, we included participants with 2 or more HVF tests, rather than the standard 5 or more. This sensitivity analysis was performed to maximize use of all available data, include patients who may have experienced vision loss over a short follow-up period, account for variations in follow-up duration, and incorporate patients with fewer follow-up visits due to DFU. Because patients with longer follow-up durations and more HVF tests could have increased opportunities for detecting disease progression, these factors—neither of which were strongly correlated—were included as matching criteria.
Cases were defined as eyes with a MD decline greater than 0.5 dB per year, determined using linear regression on MD values over the study period. Controls were eyes without an MD decline greater than 0.5 dB per year. Cases and controls were 1:1 matched based on age (±5 years), follow-up duration (±6 months), number of HVF tests (exact match), and baseline MD (±2 dB).
Within this case-control cohort, separate multivariate logistic regression models were used to assess associations between each DFU metric (percentage of DFUs, mean days of DFU, and maximum days of DFU) and HVF progression. These models were adjusted for demographics (self-reported sex and race), a binary SDOH indicator, baseline ocular measures (OCT RNFL thickness, IOP, CCT, CDR), and number of visits. Matching variables (age, baseline MD, follow-up duration, and number of HVF tests) were excluded from the models to prevent overfitting, as they were already balanced between cases and controls. Additionally, subgroup analyses by disease severity were performed with the same covariate adjustment
RESULTS
Baseline demographics and clinical characteristics of patients with 5 or more HVF tests are shown in Table 1 . A total of 1121 eyes from 600 patients were included in the primary analysis. The mean follow-up duration was 7.3 years (range: 1.3 to 10.9 years), during which participants underwent mean of 7.6 HVF tests (range: 5 to 24). The mean (SD) of baseline age was 73.7 (12.2) years, 50.2% of participants were female, 44.0% self-identified as White, and 5.8% indicated an unmet SDOH need. The mean provider-recommended follow-up interval was 166.8 days, and 53.4% of subsequent visits occurred after the recommended follow-up interval were categorized as having DFU. Upon extensive review of internal and available external medical records of all 391 patients with maximum DFU ≥180 days, 28 patient were found to have documented outside care, only 11 patients outside care involved glaucoma. Of these 11 patients, just 3 received glaucoma care elsewhere during a DFU period. No significant associations were identified between the number of cancelation types documented in the EMR (patient cancelation, patient no-show, provider cancelation, vs COVID-related cancelation) and odds of glaucoma progression.
TABLE 1
Baseline and Follow-Up Characteristics of Open Angle Glaucoma Patients and Suspects at a Tertiary Care University Hospital.
| Variable | N = 600 Patients |
|---|---|
| Baseline patient characteristics | Mean (SD) |
| Age in years, mean (SD) | 73.7 (12.2) |
| Female Sex, n (%) | 301 (50.17) |
| Self-Reported Race, n (%) | |
| White | 264 (44.00) |
| Black | 45 (7.50) |
| Asian | 218 (36.33) |
| Other | 73 (12.17) |
| Patient indicated an unmet social determinant of health need, n (%) | 35 (5.83) |
| Baseline Ocular Characteristics | N = 1121 eyes |
| IOP (mmHg), mean (SD) | 16.58 (4.47) |
| HVF MD (dB), mean (SD) | −4.56 (6.01) |
| Average OCT RNFL thickness (μm), mean (SD) | 78.65 (14.14) |
| Central corneal thickness(μm), mean (SD) | 534.69 (41.46) |
| Cup-to-disc ratio, mean (SD) | 0.67 (0.19) |
| Clinical follow-up characteristics | N = 1121 eyes |
| Study follow-up length (years): mean (SD) | 7.30 (2.12) |
| Number of HVF tests, mean (SD) | 7.57 (2.61) |
| Provider recommended follow-up interval between visits (days), mean (SD) | 166.76 (46.65) |
| Mean days of DFU: mean (SD) | 42.29 (38.51) |
| Maximum days of DFU: mean (SD) | 273.88 (227.62) |
| Percentage of visits with delayed follow-up, mean (SD) | 53.36% (20.73%) |
IOP = intraocular pressure; HVF = Humphrey Visual Field; MD = mean deviation; SD = standard deviation; OCT = optical coherence tomography; RNFL = retinal nerve fiber layer; DFU = delayed follow-up.
Participants’ demographics and clinical characteristics by glaucoma severity are shown in Supplementary Table 1. Eyes with more severe glaucoma at a baseline (i.e., baseline MD ≤−6 dB, n = 301 eyes) had a higher rate of glaucoma progression over the study period, shorter follow-up length, and shorter recommended clinical follow-up intervals. The mean (SD) provider-recommended follow-up interval was 176.2 (42.3) days for patients with baseline MD >−6 dB, and 125.9 (4.6) days for those with baseline MD ≤−6 dB ( P <.001). During follow-up period, 19.8% (99/820) of eyes with baseline MD >−6 dB and 32.8% (78/301) of eyes with baseline MD ≤−6 dB demonstrated HVF progression (i.e., an MD slope of <−0.5 dB per year).
PREDICTOR: DFU METRIC DEFINED AS PERCENTAGE OF VISITS WITH DFU
Percentage of DFU was categorized using tertiles: 0% to 45%, 45.1% to 65%, and >65% DFU. Compared to eyes in the lowest DFU tertile, those in the second (aOR: 0.78; 95% CI: 0.50-1.22; P =.29) and third (aOR: 0.83; 95% CI: 0.50-1.37; P =.47) tertiles did not have significantly different odds of HVF progression in the multivariable analysis. However, older age (per 10-year aOR: 1.34; 95% CI: 1.11-1.61; P =.002), higher CDR (per 0.1 increase aOR: 1.13 95% CI: 1.00-1.28, P =.04), shorter follow up length (per 1-year aOR: 0.81; 95% CI: 0.74-0.89), P <.001), higher number of visits (per 1 visit aOR: 1.02; 95% CI: 1.01-1.04, P <.001), and having an unmet SDOH need (aOR: 1.88; 95% CI: 1.03-3.43; P =.04) were significantly associated with higher risk of MD progression ( Table 2 ). The results were similar when using mean deviation slope as a continuous outcome in the models (Supplementary Table 3).
TABLE 2
Multivariable Analysis of the Association Between Percentage of Visits With Delayed Clinical Follow-Up and Humphrey Visual Field Mean Deviation Progression.
| Variable | Level or Increment |
N of Eyes
N = 1121 |
N of Eyes With
MD Progression (%) |
Adjusted OR (95% CI) | P -Value |
|---|---|---|---|---|---|
| Percentage of visits with DFU | 0%-45% | 367 | 85 (23.2) | REF | |
| 45%-65% | 429 | 67 (15.6) | 0.78 (0.50-1.22) | .29 | |
| >65% | 325 | 46 (14.2) | 0.83 (0.50-1.37) | .47 | |
| Age | Per 10 years | 1.34 (1.11-1.61) | .002 | ||
| Sex | Male | 553 | 101 (18.3) | REF | |
| Female | 568 | 97 (17.1) | 1.06 (0.74-1.53) | .73 | |
| Self-reported race | White | 503 | 85 (16.9) | REF | |
| Black | 80 | 15 (18.8) | 1.07 (0.51-2.21) | .86 | |
| Asian | 402 | 74 (18.4) | 1.06 (0.7-1.61) | .78 | |
| Other | 136 | 24 (17.6) | 1.00 (0.53-1.87) | .99 | |
| Average OCT RNFL thickness | Per 10 um increase | 0.88 (0.72-1.09) | .25 | ||
| HVF MD | Per 10 dB increase | 0.99 (0.95-1.02) | .40 | ||
| Intraocular pressure | Per 10 mmHg increase | 1 (0.97-1.04) | .88 | ||
| Central corneal thickness | Per 10 um increase | 0.98 (0.93-1.03) | .38 | ||
| Cup to disc ratio | Per 0.1 increase | 1.13 (1.00-1.28) | .04 | ||
| Follow up length | Per 1 year increase | 0.81 (0.74-0.89) | <.001 | ||
| Number of visits during the follow-up period | Per 1 visit increase | 1.02 (1.01-1.04) | <.001 | ||
| Patient indicated an unmet social need | No | 1053 | 178 (16.9) | REF | |
| Yes | 68 | 20 (29.4) | 1.88 (1.03-3.43) | .04 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
