Socioeconomic and Clinical Factors Associated with Keratoconus Severity and Follow-Up in a Tertiary Referral Cohort Authors: Fares Alrefai, Karanpreet S Multani, Liam D Redden, Jessica Hsueh, Kai Ding, Kamran M Riaz Published: 2026 ABSTRACT AbstractPurposeTo evaluate the severity of keratoconus (KCN) at presentation in self-identified minority patients, identify risk factors for progression and surgery, and assess loss to follow-up.Material and MethodsPentacam tomography records (2014–2021) from a single academic center were screened for KCN using BAD-D ≥1.6 and Kmax ≥47D. Manual chart review confirmed diagnosis, determined severity, and assessed progression and surgical intervention within two years. Demographic, socioeconomic, and clinical parameters were recorded. Logistic regression and generalized linear mixed models were used to evaluate associations, with covariates selected from univariate analysis; final models were determined by backward elimination. The primary outcome was KCN severity at presentation; secondary outcomes were progression, surgery within two years, and loss to follow-up.ResultsOf 21,157 charts reviewed, 797 patients had KCN; 479 were male (60.1%) and 458 (57.5%) identified as racial/ethnic minorities. Mean age at presentation was 41.7 ± 16.5 years. Severe disease was present in 635 (81.8%) patients. Patients with connective tissue disease were less likely to present with severe KCN (p = 0.0015). Younger age predicted surgical recommendation within two years (p = 0.008). Patients recommended for surgery were more likely to be younger (p = 0.008) and to speak a primary language other than English (p = 0.006). Two-year loss to follow-up (LTFU) was high (65.1%). Smokers were associated with a greater LTFU (p = 0.0475). However, patients with eczema (p = 0.0388) or those with other ocular diseases (p = 0.0058) were less likely to be LTFU.ConclusionMarkers of socioeconomic disadvantage cluster with worsened disease burden but only select factors remained independently associated after adjustment.Keywords: keratoconus, socioeconomic factors, care disparity, disease burden, patient demographics Introduction Keratoconus (KCN) is the most common type of corneal ectasia and is a bilateral, typically asymmetric disorder characterized by progressive thinning and steepening of the cornea, resulting in irregular astigmatism and vision impairment.1,2 The worldwide prevalence of KCN varies and is affected by environmental, genetic, and ethnic factors, making it a complex disease to measure consistently.1 Environmental factors such as eye rubbing and nocturnal ocular compression, as well as genetic factors including family history, ethnicity, and consanguinity, have all been associated with the development of the disease.3,4 Recent international consensus statements emphasize a paradigm shift in KCN care from detecting clinically manifest disease toward identifying ectasia susceptibility, facilitating earlier diagnosis, longitudinal monitoring, and preventive intervention before irreversible vision loss occurs.5,6 KCN can place a significant burden on both patients and healthcare systems due to its potential to cause severe vision loss and the expensive treatments involved, such as corneal cross-linking (CXL) and keratoplasty. Large epidemiologic studies have shown disparities in eye care utilization and a higher risk for losing sight across racial and ethnic minority groups for diseases such as glaucoma, diabetic retinopathy, and cataracts.7,8 However, there is a relative paucity of literature examining the impact of socioeconomic factors on the severity and progression of KCN. Several retrospective cohort studies have reported that socioeconomic disadvantage, non-commercial insurance, and minority status are associated with more severe KCN at presentation, higher rates of visual impairment, and increased likelihood of surgical intervention.1,9–12 Consistent with the First and Second Global Consensus on Keratoconus and Ectatic Diseases, which emphasize early diagnosis, longitudinal monitoring, and timely intervention as central principles of KCN management, understanding barriers that contribute to delayed presentation and interruptions in care is of increasing clinical importance.5,6 Identifying these disparities is essential, as early diagnosis and timely management of KCN, especially in the current era of less invasive treatments, can prevent severe visual loss and the need for more invasive procedures, such as keratoplasty. This study extends existing literature by examining the impact of underexplored socioeconomic factors, including primary language, education attainment, and community-level life expectancy, on KCN severity and care utilization. Preliminary findings from this cohort were previously presented as an abstract. The present study expands upon that work through comprehensive multivariable analyses, additional socioeconomic variables, and a full evaluation of clinical outcomes.13 This retrospective cohort study aimed to examine the association between socioeconomic factors and the severity of KCN at presentation among a diverse population at a single academic referral center over eight years. By analyzing the demographic, socioeconomic, and clinical characteristics of patients, we sought to identify predictors of severe disease at presentation, its progression over time, and access to recommended surgical treatments. Understanding these factors can help develop strategies to reduce disparities in KCN care and improve outcomes for vulnerable and underserved populations. Methods This study was conducted per the ethical standards of the Helsinki Declaration and the Health Insurance Portability and Accountability Act (HIPAA). Institutional Review Board (IRB:13118) approval was obtained from the University of Oklahoma Health Sciences Center. A retrospective chart review of new patient encounters at a single academic center with tomographic images (Pentacam; Oculus Optikgeräte GmbH, Wetzlar, Germany) from 2014 through 2021 was conducted. Patients were initially screened using tomographic criteria (BAD-D ≥1.6 and Kmax ≥47 D). KCN diagnoses were subsequently confirmed by manual review of ICD-9 or ICD-10 diagnostic codes together with the corresponding clinical documentation. Longitudinal analyses included data from both eyes when applicable, and generalized linear mixed models accounted for inter-eye correlation by specifying eye nested within patient as a random effect. Patients with documented corneal scarring from other causes (eg, trauma or infection), other corneal ectasias (eg, pellucid marginal degeneration), prior corneal refractive surgery, or prior KCN surgery (eg, corneal transplantation or CXL) were excluded. Patients with concurrent ocular comorbidities (eg, glaucoma or diabetic retinopathy) were not excluded unless the condition precluded accurate assessment of KCN. Baseline severity (steep K2/Kmax) and year of presentation were included in generalized linear mixed models (GLMMs) to account for disease severity and temporal confounding (CXL adoption, pandemic). Social demographics (age, gender, race, primary language), geographical information (zip code), and insurance data were collected. Race and ethnicity were self-reported and categorized according to NIH/US Census Bureau designations: American Indian or Alaska Native, Asian, Black or African American, Native Hawaiian or Other Pacific Islander, White, and Hispanic/Latino. “Other” and “two or more races” were also recorded. Zip codes were used as a proxy for household income, educational status, and life expectancy using 2022 US Census Data. Of note, life expectancy analyses reflect a reduced sample size and should be interpreted as exploratory. Analyses including this variable were performed using complete-case analysis; therefore, patients with missing life expectancy data were excluded from models containing this covariate. Patient visits were assessed at presentation and four intervals of approximately 6 months each for 24 months. We defined severe KCN as K2 ≥52 D, consistent with CLEK and other large cohort studies that have used similar thresholds to classify advanced disease.14–16 Progression was defined as ≥1 D increase in steep K, in line with CLEK methodology and subsequent literature indicating that this cutoff exceeds tomographic repeatability and represents clinically meaningful change.16–18 We used K2 to define severity because it is less susceptible to local cone irregularities or measurement noise compared to Kmax, making it a more stable parameter for defining severity and tracking progression across visits. Kmax was more suitable for our inclusion criteria as it better identifies eyes likely to have clinically manifest KCN. This threshold has been used in the CLEK study and subsequent literature; although measurement repeatability introduces noise, ≥1 D is a clinically accepted progression threshold.14 Patients who were recommended and received surgery at any point in the initial two years were then followed for an additional two years from the date of surgery. Loss to follow-up (LTFU) was determined based on the initial clinic exam date at which Pentacam data were obtained. Surgeries included CXL, intracorneal ring segment (ICRS; CorneaGen, Inc., Seattle, Washington, USA) implant, or corneal transplant. The primary outcome was disease severity at presentation. The secondary outcomes were disease progression, obtaining recommended surgery, and LTFU. All outcomes are binary. Descriptive statistics (mean, standard deviation, count, percentage) were used to summarize the data. Binary outcomes, including disease severity at presentation and LTFU, were modeled using logistic regression. When sparse data prevented convergence with standard maximum likelihood estimation, Firth’s penalized likelihood method was used to reduce small-sample bias. Odds ratios (OR) with 95% confidence intervals (CI) were reported. Multivariable logistic regression models were built by including covariates with p < 0.20 in univariate analysis; final models were selected through backward elimination for simplicity. Effect sizes and CIs are emphasized over p-values, reflecting the exploratory nature of this study. All multivariable models were adjusted for age to account for its strong influence on severity, progression, and surgery risk. For outcomes with a clustered and longitudinal structure (disease progression and surgery within two years), GLMMs with a logit link were used. Patient and eye (nested within patient) were specified as random effects to account for intra-patient correlation. Univariate GLMMs were fit for each baseline or socioeconomic variable individually. Multivariable GLMMs included variables with p < 0.20 from univariate models, categorical time points (baseline, 6, 12, 18, and 24 months), and two-way interactions with time. No significant interactions were identified. Final GLMMs were selected through backward elimination. Model assumptions were checked, and significance was defined as p < 0.05. Given the interrelated nature of demographic and socioeconomic variables, multivariable analyses are emphasized as the primary findings; univariate results are presented for completeness. Time-to-event analysis for retention in care was visualized with a Kaplan–Meier curve, and odds ratios from multivariable models for LTFU were displayed as a forest plot (Supplementary Figures). The SAS software (v9.4) was used for all analyses. Results Of 21,157 patient charts reviewed, 797 patients were confirmed to have KCN at presentation. Baseline characteristics of all KCN patients studied are presented in Supplementary Table 1. The mean age at presentation was 41.7 ± 16.5 years. A total of 60.1% (n = 479) were male. By race/ethnicity, 42.5% (n = 339) were White and 57.5% (n = 458) identified as racial/ethnic minorities. On average, patients traveled 61.6 miles to the clinic, and the median income by residential zip code was $65,000 per year. Among 255 surgeries performed, 46.7% (n = 119) were CXL, 11.7% (n = 30) were INTACS, 0.2% (n = 5) were DALK, and 41.4% (n = 101) were PKP. Most patients, 94.7% (n = 752), reported English as their primary language, with only 5.3% (n = 42) reporting a primary language other than English (PLOTE). Insurance status was distributed as follows: 57.8% (n = 461) with commercial insurance, 18.4% (n = 147) with Medicare, 9.8% (n = 78) with Medicaid, and 14.0% (n = 111) on financial assistance, uninsured, or incarcerated. A family history of KCN was reported in 6.5% (n = 52). Medical comorbidities included atopy in 30.4% (n = 242), diabetes in 12.8% (n = 102), connective tissue disease in 5.0% (n = 40), hormonal imbalance in 6.5% (n = 52), and dry eye in 22.4% (n = 178). At presentation, 81.8% (n = 635) had severe KCN disease. Over 24 months, 65.1% (n = 519) of patients were lost to follow-up. Table 1 summarizes predictors of KCN severity at presentation. The average age of patients who presented with severe disease was 40.3±15.9 years, whereas that of patients with non-severe disease was 46.5±18.2 years. On multivariable logistic regression, only connective tissue disease was found to be significantly associated with KCN severity (OR = 0.192, p = 0.0015). Table 1Univariate and Multivariate Analysis of Severe Disease at Presentation*VariableUnivariate AnalysisMultivariate AnalysisOR (95% CI)p-valueOR (95% CI)p-valueGender (F vs M)0.746 (0.516,1.078)0.1184––Race0.0382(**)–– Asian vs White2.245 (0.276, 18.250) Black vs White1.713 (1.068, 2.748) Hispanic vs White2.077 (1.140, 3.784) Native American/Alaskan vs White1.291 (0.620, 2.686) Other/unknown vs White0.748 (0.391, 1.432)Age at presentation (1 year increase)0.978 (0.967,0.989)<0.0001(**)––Distance (1 mile increase)1 (0.998,1.002)0.9911––Median income ($1000 increase)0.995 (0.986,1.003)0.1909––Percent of population age 25+ with bachelors or higher (1 percentage point increase)0.992 (0.979,1.006)0.2587––Percent of population age 25+ with completed high school or higher education (1 percentage point increase)0.973 (0.945,1.001)0.0549 (***)––Life expectancy in zip code (1 year increase)0.882 (0.802,0.971)0.0106(**)0.888 (0.806,0.978)0.0157 (**)Insurance0.0075(**)–– DOC vs commercial2.472 (0.567,10.774) Medicaid vs commercial1.681 (0.832,3.397) Medicare vs commercial0.763 (0.488,1.193) Uninsured/PFAP vs commercial3.419 (1.446,8.087)Language (PLOTE vs English)4.696 (1.121,19.663)0.0343(**)––Job status0.0928–– Retired vs employed0.854 (0.325,2.243) Student vs employed0.759 (0.322,1.79) Unemployed/Disabled vs employed4.138 (1.174,14.589)Smoker (yes vs no)1.323 (0.74,2.366)0.3456––Family history (yes vs no)1.139 (0.522,2.486)0.7440––Allergies (yes vs no)0.959 (0.598,1.537)0.8621––Asthma (yes vs no)1.051 (0.636,1.734)0.8473––Eczema (yes vs no)0.876 (0.375,2.049)0.7608––Diabetes (yes vs no)1.165 (0.659,2.059)0.5986––Connective tissue disease (yes vs no)0.26 (0.134,0.503)0.0001(**)0.192 (0.069,0.532)0.0015 (**)Down syndrome (yes vs no)0.658 (0.131,3.292)0.6101––Hormonal imbalance (yes vs no)0.486 (0.258,0.917)0.0259(**)––Sleep apnea (yes vs no)5.641 (0.296,107.506)0.2500––Dry eye (yes vs no)0.972 (0.628,1.504)0.8984––Other ocular disease (yes vs no)0.729 (0.495,1.072)0.1081––Atopy (yes vs no)0.875 (0.592,1.294)0.5039––Notes: *Multivariate logistic regression analysis included variables with a p-value < 0.2 from univariate analysis. For sparse data, Firth’s bias-reducing penalized likelihood method was used. The final model was reached by backward selection. **Statistically significant values. ***Borderline statistically significant values.Abbreviation: DOC, department of corrections; PLOTE, Primary language other than English; PFAP, patient financial assistance programs. Univariate and Multivariate Analysis of Severe Disease at Presentation* Notes: *Multivariate logistic regression analysis included variables with a p-value < 0.2 from univariate analysis. For sparse data, Firth’s bias-reducing penalized likelihood method was used. The final model was reached by backward selection. **Statistically significant values. ***Borderline statistically significant values. Abbreviation: DOC, department of corrections; PLOTE, Primary language other than English; PFAP, patient financial assistance programs. Table 2 presents risk factors associated with KCN disease progression within 24 months. On multivariable analysis, having another ocular disease before KCN diagnosis was a predictor for disease progression (OR = 0.576, p = 0.018). Table 2Univariate and Multivariate Analysis of Disease Progression*VariableUnivariate Analysis†Multivariate Analysis‡OR (95% CI)p-valueOR (95% CI)p-valueGender (F vs M)0.894 (0.604, 1.324)0.576––Race0.511 Asian vs white0.458 (0.035, 6.008) Black vs white1.580 (0.980, 2.548) Hispanic vs white1.087 (0.617, 1.918) Native American/Alaskan vs white1.145 (0.530, 2.474) Other/unknown vs white1.175 (0.522, 2.645)Age at presentation (1 year increase)0.985 (0.973, 0.998)0.029(**)––Distance (1 mile increase)1.000 (0.997, 1.002)0.853––Median income ($1000 increase)0.997 (0.988, 1.006)0.547––Percent of population age 25+ with bachelors or higher (1 percentage point increase)0.992 (0.977, 1.006)0.259––Percent of population age 25+ with high school or higher (1 percentage point increase)0.992 (0.967, 1.017)0.509––Life expectancy in zip code (1 year increase)0.980 (0.898, 1.070)0.652––Insurance0.623–– DOC vs commercial0.903 (0.277, 2.947) Medicaid vs commercial1.498 (0.805, 2.787) Medicare vs commercial0.974 (0.546, 1.740) Uninsured/PFAP vs commercial1.406 (0.730, 2.709)Language (English vs PLOTE)1.283 (0.550, 2.995)0.564––Job status0.120–– Disabled vs employed0.703 (0.101, 4.868) Retired vs employed0.510 (0.124, 2.102) Student vs employed2.632 (1.102, 6.287) Unemployed vs employed1.535 (0.580, 4.063)Smoker (yes vs no)0.632 (0.303, 1.315)0.219––Family history (yes vs no)1.177 (0.555, 2.499)0.670––Allergies (yes vs no)1.307 (0.828, 2.064)0.249––Asthma (yes vs no)1.198 (0.723, 1.984)0.483––Eczema (yes vs no)1.201 (0.464, 3.108)0.705––Diabetes (yes vs no)0.500 (0.237, 1.054)0.068––Connective tissue disease (yes vs no)0.279 (0.071, 1.090)0.066––Down syndrome (yes vs no)0.100 (0.005, 1.870)0.123––Hormonal imbalance (yes vs no)0.548 (0.233, 1.287)0.167––Sleep apnea (yes vs no)0.536 (0.107, 2.676)0.446––Dry eye (yes vs no)0.706 (0.443, 1.127)0.145––Other ocular disease (yes vs no)0.576 (0.365, 0.911)0.018(**)0.576 (0.365, 0.909)0.018(**)Atopy (yes vs no)1.400 (0.939, 2.087)0.098––Notes: *A generalized linear mixed model (logit link) was used for clustered longitudinal data. Both patient and eye nested within patient were treated as a random effect. **Statistically significant values. †Univariate analysis considered each variable separately and also included baseline steep K, (categorical) time point (months 6, 12, 18, and 24), and interactions (time × variable, time × baseline steep K). No significant interactions were found. The final model was reached by backward selection. ‡Multivariate analysis included variables with p-value < 0.2 from univariate analysis and also included baseline steep K, (categorical) time point (months 6, 12, 18, and 24), and interactions (time × each included variable, time × baseline steep K). No significant interactions were found. Baseline steep K was not significant in multivariate analysis. The final model was reached by backward selection.Abbreviations: DOC, department of corrections; PLOTE, Primary language other than English; PFAP, patient financial assistance programs. Univariate and Multivariate Analysis of Disease Progression* Notes: *A generalized linear mixed model (logit link) was used for clustered longitudinal data. Both patient and eye nested within patient were treated as a random effect. **Statistically significant values. †Univariate analysis considered each variable separately and also included baseline steep K, (categorical) time point (months 6, 12, 18, and 24), and interactions (time × variable, time × baseline steep K). No significant interactions were found. The final model was reached by backward selection. ‡Multivariate analysis included variables with p-value < 0.2 from univariate analysis and also included baseline steep K, (categorical) time point (months 6, 12, 18, and 24), and interactions (time × each included variable, time × baseline steep K). No significant interactions were found. Baseline steep K was not significant in multivariate analysis. The final model was reached by backward selection. Abbreviations: DOC, department of corrections; PLOTE, Primary language other than English; PFAP, patient financial assistance programs. Table 3 identifies factors associated with patients requiring ocular surgery within 2 years of initial presentation. On multivariable analysis, younger patients (OR = 0.989, p = 0.008), PLOTE (OR = 2.958, p = 0.006), and patients with Steep K increases >1 (OR = 1.030, p < 0.0001) were more likely to require surgery. Surgery reflects both disease biology and system-level access factors and should not be interpreted solely as a marker of disease aggressiveness. Table 3Univariate and Multivariate Analysis of Required Surgery*VariableUnivariate Analysis†Multivariate Analysis‡OR (95% CI)P-valueOR (95% CI)P-valueGender (F vs M)1.165 (0.891, 1.523)0.264––Race0.449 Asian vs white1.565 (0.496, 4.937) Black vs white1.157 (0.832, 1.608) Hispanic vs white0.985 (0.656, 1.477) Native American/Alaskan vs white1.203 (0.710, 2.036) Other/unknown vs white1.647 (0.988, 2.746)Age at presentation (1 year increase)0.991 (0.983, 0.999)0.023(**)0.989 (0.981, 0.997)0.008(**)Distance (1 mile increase)1.000 (0.999, 1.001)0.941––Median income ($1000 increase)0.997 (0.991, 1.004)0.413––Percent of population age 25+ with bachelors or higher (1 percentage point increase)0.994 (0.984, 1.004)0.257––Percent of population age 25+ with high school or higher (1 percentage point increase)0.995 (0.978, 1.012)0.532––Life expectancy in zip code (1 year increase)1.028 (0.967, 1.093)0.379––Insurance0.069–– DOC vs commercial0.382 (0.133, 1.100) Medicaid vs commercial1.229 (0.805, 1.878) Medicare vs commercial0.967 (0.682, 1.371) Uninsured/PFAP vs commercial0.606 (0.371, 0.988)Language (English vs PLOTE)2.652 (1.223, 5.751)0.014(**)2.958 (1.360, 6.433)0.006(**)Job status0.188–– Disabled vs employed1.747 (0.720, 4.240) Retired vs employed1.587 (0.808, 3.115) Student vs employed1.611 (0.861, 3.014) Unemployed vs employed0.740 (0.354, 1.546)Smoker (yes vs no)0.942 (0.630, 1.408)0.771––Family history (yes vs no)1.487 (0.902,2.452)0.120––Allergies (yes vs no)0.959 (0.682,1.349)0.811––Asthma (yes vs no)1.207 (0.856,1.701)0.283––Eczema (yes vs no)1.124 (0.6,2.105)0.715––Diabetes (yes vs no)0.871 (0.579,1.31)0.507––Connective tissue disease (yes vs no)1.162 (0.647,2.087)0.615––Down syndrome (yes vs no)0.229 (0.027,1.946)0.177––Hormonal imbalance (yes vs no)0.636 (0.348,1.163)0.142––Sleep apnea (yes vs no)1.294 (0.486,3.447)0.607––Dry eye (yes vs no)0.708 (0.508,0.986)0.041(**)––Other ocular disease (yes vs no)0.764 (0.57,1.024)0.072––Atopy (yes vs no)1.124 (0.847,1.491)0.419––Steep K (1 unit increase)1.030 (1.020, 1.041)<0.00011.030 (1.019, 1.041)<0.0001Notes: *A generalized linear mixed model (logit link) was used for clustered longitudinal data. Both patient and eye nested within patient were treated as a random effect. **Statistically significant values. †Univariate analysis considered each variable separately and included baseline steep K, (categorical) time point (baseline, months 6, 12, 18, and 24), and interaction (time × variable, time × baseline steep K). No significant interaction was found. The final model was reached by backward selection. ‡Multivariate analysis included variables with P-value < 0.2 from univariate analysis and also included baseline steep K, (categorical) time point (baseline, months 6, 12, 18, and 24), and interactions (time × each included variable, time × baseline steep K). No significant interactions were found. The final model was reached by backward selection.Abbreviations: DOC, department of corrections; PLOTE, Primary language other than English; PFAP, patient financial assistance programs. Univariate and Multivariate Analysis of Required Surgery* Notes: *A generalized linear mixed model (logit link) was used for clustered longitudinal data. Both patient and eye nested within patient were treated as a random effect. **Statistically significant values. †Univariate analysis considered each variable separately and included baseline steep K, (categorical) time point (baseline, months 6, 12, 18, and 24), and interaction (time × variable, time × baseline steep K). No significant interaction was found. The final model was reached by backward selection. ‡Multivariate analysis included variables with P-value < 0.2 from univariate analysis and also included baseline steep K, (categorical) time point (baseline, months 6, 12, 18, and 24), and interactions (time × each included variable, time × baseline steep K). No significant interactions were found. The final model was reached by backward selection. Abbreviations: DOC, department of corrections; PLOTE, Primary language other than English; PFAP, patient financial assistance programs. Table 4 presents predictors of loss to follow-up. On multivariable analysis, smokers were significantly more likely to be lost to follow-up (OR = 1.596, p = 0.0475). Eczema (OR = 0.484, p = 0.0388) and another ocular disease (OR = 0.642, p = 0.0058) were associated with a lower risk of loss to follow-up. A Kaplan–Meier-style survival curve illustrating retention in care over 24 months is shown in Supplementary Figure 1. A multivariable forest plot of predictors for LTFU is provided in Supplementary Figure 2. Table 4Univariate and Multivariate Analysis of Loss to Follow-Up (Yes vs No) *VariableUnivariate AnalysisMultivariate AnalysisOR (95% CI)p-valueOR (95% CI)p-valueGender (F vs M)0.932 (0.692,1.254)0.6403––Race0.5578 Asian vs white1.915 (0.392, 9.360) Black vs white0.925 (0.646, 1.325) Hispanic vs white1.283 (0.826, 1.993) Native American/Alaskan vs white0.786 (0.441, 1.400) Other/unknown vs white1.248 (0.687, 2.268)Age at presentation (1 year increase)0.992 (0.984,1.001)0.0898––Distance (1 mile increase)0.999 (0.998,1.001)0.4366––Median income ($1000 increase)0.997 (0.99,1.004)0.3832––Percent of population age 25+ with bachelors or higher (1 percentage point increase)0.999 (0.989,1.01)0.9115––Percent of population age 25+ with high school or higher (1 percentage point increase)0.989 (0.97,1.009)0.2962––Life expectancy in zip code (1 year increase)0.963 (0.9,1.031)0.2841––Insurance0.0679–– DOC vs commercial1.377 (0.528,3.589) Medicaid vs commercial0.742 (0.455,1.213) Medicare vs commercial0.728 (0.498,1.065) Uninsured/PFAP vs commercial1.572 (0.936,2.641)Language (PLOTE vs English)1.214 (0.62,2.375)0.5714––Job status0.2047–– Retired vs employed0.881 (0.403,1.926) Student vs employed1.695 (0.799,3.594) Unemployed/Disabled vs employed1.754 (0.901,3.417)Smoker (yes vs no)1.476 (0.936,2.327)0.09351.596 (1.005,2.534)0.0475(**)Family history (yes vs no)0.849 (0.476,1.515)0.5803––Allergies (yes vs no)0.905 (0.622,1.317)0.6010––Asthma (yes vs no)1.268 (0.846,1.9)0.2496––Eczema (yes vs no)0.49 (0.248,0.967)0.0397(**)0.484 (0.243,0.963)0.0388(**)Diabetes (yes vs no)0.89 (0.579,1.37)0.5972––Connective tissue disease (yes vs no)0.997 (0.512,1.941)0.9918––Down syndrome (yes vs no)0.893 (0.212,3.767)0.8780––Hormonal imbalance (yes vs no)1.113 (0.611,2.024)0.7270––Sleep apnea (yes vs no)0.749 (0.236,2.383)0.6251––Dry eye (yes vs no)0.759 (0.538,1.07)0.1155––Other ocular disease (yes vs no)0.649 (0.475,0.887)0.0066(**)0.642 (0.468,0.879)0.0058(**)Atopy (yes vs no)0.935 (0.682,1.282)0.6757––Notes: *Multivariate logistic regression analysis included variables with p-value < 0.2 from univariate analysis. For sparse data, Firth’s bias-reducing penalized likelihood method was used. Final model was reached by backward selection. **Statistically significant values.Abbreviations: DOC, department of corrections; PLOTE, Primary language other than English; PFAP, patient financial assistance programs. Univariate and Multivariate Analysis of Loss to Follow-Up (Yes vs No) * Notes: *Multivariate logistic regression analysis included variables with p-value < 0.2 from univariate analysis. For sparse data, Firth’s bias-reducing penalized likelihood method was used. Final model was reached by backward selection. **Statistically significant values. Abbreviations: DOC, department of corrections; PLOTE, Primary language other than English; PFAP, patient financial assistance programs. Discussion Our retrospective cohort study identified multiple socioeconomic and demographic factors associated with more severe KCN at presentation. On univariate analysis, patients presenting with severe disease were more likely to be non-White, younger, have lower educational attainment, live in zip codes with lower life expectancy, lack commercial insurance, and report a PLOTE. These findings align with disparities observed across other ophthalmic conditions and reinforce that social determinants of health may similarly influence KCN care.12,19 However, many of these associations were attenuated in multivariable analyses, suggesting that several socioeconomic indicators may represent overlapping structural and demographic factors rather than independent predictors. Accordingly, the adjusted analyses should be considered the primary findings of this exploratory study, while univariate associations are best interpreted as hypothesis-generating. Notably, more than 80% of patients in our cohort presented with severe KCN at their initial visit. This proportion is substantially higher than that reported in community-based and registry cohorts (~40–55%) and likely reflects referral and ascertainment effects inherent to a tertiary-care academic center cornea practice in the Southwestern United States, as well as our tomographic screening thresholds (BAD-D ≥1.6; Kmax ≥47 D).20 These thresholds were intentionally selected to improve diagnostic certainty for clinically manifest KCN but likely excluded milder or forme fruste disease. These design features enrich for advanced phenotypes and may inflate estimates of severity at presentation relative to the broader KCN population. This finding likely reflects delayed diagnosis due to limited access to subspecialty eye care or reduced awareness of early disease, although referral bias remains an important contributing factor. Consequently, our findings may not generalize to community-based or population-screened KCN cohorts, where milder disease phenotypes are more commonly represented. Younger age was consistently associated with greater disease severity at presentation, higher likelihood of progression, and increased need for surgical intervention, consistent with prior studies demonstrating more aggressive disease in younger patients (Tables 1–3).18,19 In contrast, male sex was not associated with worse outcomes, differing from reports describing greater disease burden among men.21,22 Racial and ethnic disparities were evident at presentation, with Black and Hispanic patients more likely to present with severe disease on univariate analysis, consistent with prior studies examining socioeconomic influences on KCN diagnosis and treatment.9,23–25 However, race was not independently associated with KCN progression or surgery once patients entered follow-up, suggesting that observed racial disparities may reflect differences in access to care, referral patterns, or delayed diagnosis rather than intrinsic differences in disease biology. This attenuation after adjustment underscores the importance of structural and social factors as mediators of observed disparities. Few variables independently predicted progression, likely reflecting limited power in the younger, higher-risk subgroup and high attrition. Because socioeconomic variables such as insurance status, language, education, and race/ethnicity are highly interrelated, residual confounding and collinearity may also have influenced the stability of these associations. Although KCN has a well-established genetic component, self-reported family history was not significantly associated with disease severity, progression, or surgical intervention in our cohort. This finding should be interpreted cautiously, as family history may have been underreported or incompletely documented in the medical record, and some affected relatives may have remained undiagnosed. Additionally, the relatively small proportion of patients reporting a positive family history may have limited statistical power to detect modest associations. Socioeconomic status was further explored using zip code–level census data as proxies for income, education, and life expectancy. Lower educational attainment was associated with more severe disease at presentation but did not affect disease progression or the need for surgery, suggesting that delayed engagement with eye care due to reduced health literacy may play a greater role than underlying disease differences. This finding is consistent with prior work showing inconsistent associations between income, education, and KCN burden.25 Importantly, these measures were derived from area-level census data and may not accurately reflect individual patient socioeconomic circumstances, raising the possibility of ecological fallacy. Distance traveled was not associated with severity or progression, suggesting that geographic distance alone did not preclude attendance at subspecialty visits once referral occurred. Insurance status and language barriers were also associated with disease severity and access to treatment. Patients without commercial insurance were more likely to present with severe disease, likely reflecting financial barriers to timely ophthalmologic evaluation. Additionally, PLOTE patients were more likely to present with severe disease and to undergo surgery within two years of follow-up. Language barriers may contribute to delayed diagnosis, impaired communication, and challenges in navigating complex care pathways.26 These findings should be interpreted cautiously given the relatively small PLOTE subgroup and wide confidence intervals. Although baseline disease severity was included in surgical models to account for severity-driven treatment decisions, these observations are best considered exploratory and hypothesis-generating rather than definitive evidence of an independent language-related disparity. Loss to follow-up (LTFU) was a significant concern, with more than 65% of patients failing to return within 2 years. This attrition rate far exceeds those reported in post-CXL cohorts and highlights systemic barriers to longitudinal KCN care. Insurance status was not an independent predictor of LTFU, suggesting that other unmeasured factors (eg, transportation challenges, competing socioeconomic priorities, or health system disruptions during the COVID-19 pandemic, other unmeasured factors) may play important roles. High attrition is particularly concerning in KCN, where timely intervention can prevent irreversible vision loss. Additionally, patients lost to follow-up may differ systematically from retained patients, raising the possibility of informative censoring and survivorship bias in analyses of progression and surgical outcomes. Consequently, estimates of progression and surgical intervention among retained patients may not fully reflect outcomes across the entire study cohort. Several ocular and systemic comorbidities warrant brief discussion. Patients with connective tissue disease (CTD) were less likely to present with severe disease, an unexpected finding that likely reflects surveillance bias. Because patients with systemic autoimmune or inflammatory conditions are more likely to be already integrated into the healthcare system and subject to regular monitoring, they have a higher probability of being referred to subspecialists earlier than the general population. Consequently, their KCN is detected at an earlier stage, which might explain this observed effect rather than an actual protective biological mechanism.27,28 Patients with other ocular comorbidities showed mixed associations with KCN progression and surgery, likely reflecting confounding by age and more frequent clinical monitoring. Smoking and eczema were not associated with worse outcomes, consistent with reports suggesting potential protective effects of smoking via collagen crosslinking and the complex role of atopy in KCN pathogenesis.29 These findings should be interpreted cautiously and are likely influenced by confounding and limited subgroup sizes. In particular, unexpected findings involving connective tissue disease, smoking, and eczema should not be interpreted as protective effects, as they may reflect surveillance bias, age-related confounding, or limited statistical power. Our findings also have important public health implications. Given the high proportion of patients presenting with advanced disease and the substantial loss to follow-up observed in this cohort, efforts to improve patient education, facilitate timely referral to corneal specialists, and promote sustained engagement in longitudinal eye care may help reduce preventable vision loss. As contemporary KCN management increasingly emphasizes earlier diagnosis and preventive intervention, public awareness initiatives may complement clinical efforts by increasing recognition of disease symptoms, modifiable risk factors such as eye rubbing, and the importance of longitudinal follow-up. Established educational organizations, including the National Keratoconus Foundation (NKCF) and Violet June, provide resources that support patient education and disease awareness.30 Accordingly, as diagnostic paradigms continue to shift toward earlier detection of ectasia susceptibility and preventive management, reducing socioeconomic barriers to timely evaluation and longitudinal follow-up may become increasingly important for optimizing visual outcomes.6 Future prospective studies should evaluate whether targeted educational interventions and improved care coordination can reduce disparities in KCN presentation and long-term management. This study has several limitations. As a retrospective analysis conducted at a single academic referral center in Oklahoma, a region in the United States characterized by substantial rural populations, socioeconomic heterogeneity, and documented health disparities, generalizability may be limited. While this may limit generalizability across settings (including international contexts), it also offers a useful lens on KCN care in regions where structural barriers to specialty eye care are prevalent. Socioeconomic variables derived from zip code–level census data may not accurately reflect individual patient circumstances, raising the possibility of ecological fallacy. More granular measures of deprivation were unavailable. Key variables, such as eye rubbing and sleeping position were not consistently documented and were unavailable for analysis. Future prospective studies are needed to better understand the interplay between these mechanical factors and the socioeconomic disparities identified here. Definitions of severe disease and progression vary across studies; although our thresholds are consistent with CLEK and subsequent literature, alternative metrics may yield different results.15–19 Inconsistent follow-up intervals and high rates of LTFU further limit interpretation of long-term outcomes. A complete-case analysis of variables with substantial missingness, including life expectancy, may also introduce bias. Furthermore, the use of univariate screening followed by backward elimination may have contributed to model instability, particularly given the correlated nature of several socioeconomic predictors. Conclusions This study identified associations between socioeconomic and structural factors and more severe KCN at presentation, as well as differential access to surgical care, within a tertiary referral population. Disparities appeared most pronounced at initial presentation, although these findings should be interpreted cautiously given the referral-enriched nature of the cohort, substantial loss to follow-up, and attenuation of several associations after multivariable adjustment. Accordingly, these findings should be viewed as exploratory and hypothesis-generating rather than evidence of causal relationships. While our study was limited to a single geographic location in the United States, similar disparities may exist in other underserved or rural settings beyond our institution’s service region, both nationally and internationally. An underexplored finding is the association between PLOTE and both KCN disease severity and surgical intervention, highlighting language as a potentially modifiable barrier to equitable care. While further studies are needed with larger sample sizes, our study supports consideration of multilingual patient education materials, interpreter services, and culturally appropriate outreach to facilitate earlier diagnosis and sustained engagement in care. Additionally, community-level life expectancy may serve as a surrogate marker of cumulative socioeconomic disadvantage relevant to KCN outcomes. Recognizing and addressing these social determinants may help optimize treatment planning, improve retention in care, and reduce disparities in KCN-related vision loss. Funding Statement The authors have no direct public or private support to disclose. The authors’ study institution is supported in part by an unrestricted grant from Research to Prevent Blindness, Inc., NY, NY, USA. This entity had no direct input in the study or preparation of this manuscript. Data Sharing Statement The data that support the findings of this study are available on request from the corresponding author, KR. The data are not publicly available due to their containing information that could compromise the privacy of research participants. Ethics Approval and Informed Consent This study was approved by the Institutional Review Board of the University of Oklahoma Health Sciences Center (IRB #13118), with a waiver of informed consent due to its retrospective nature. This study was conducted per the ethical standards of the Helsinki Declaration and the Health Insurance Portability and Accountability Act (HIPAA). Disclosure The authors declare no competing interests. References Associated Data Data Availability Statement The data that support the findings of this study are available on request from the corresponding author, KR. The data are not publicly available due to their containing information that could compromise the privacy of research participants.