Investigating the Accuracy and Completeness of an Artificial Intelligence Large Language Model About Uveitis: An Evaluation of ChatGPT

PURPOSE

To assess the accuracy and completeness of ChatGPT-generated answers regarding uveitis description, prevention, treatment, and prognosis.

METHODS

Thirty-two uveitis-related questions were generated by a uveitis specialist and inputted into ChatGPT 3.5. Answers were compiled into a survey and were reviewed by five uveitis specialists using standardized Likert scales of accuracy and completeness.

RESULTS

In total, the median accuracy score for all the uveitis questions (n = 32) was 4.00 (between "more correct than incorrect" and "nearly all correct"), and the median completeness score was 2.00 ("adequate, addresses all aspects of the question and provides the minimum amount of information required to be considered complete"). The interrater variability assessment had a total kappa value of 0.0278 for accuracy and 0.0847 for completeness.

CONCLUSION

ChatGPT can provide relatively high accuracy responses for various questions related to uveitis; however, the answers it provides are incomplete, with some inaccuracies. Its utility in providing medical information requires further validation and development prior to serving as a source of uveitis information for patients.

Quantitative Ocular Surface Changes in Patients Undergoing Immune Checkpoint Inhibitor Therapy

PURPOSE

To describe the clinical course and evaluate treatment of ocular surface changes in patients receiving immune checkpoint inhibitor (ICI) therapy.

METHODS

Multiple markers of ocular surface dryness were evaluated in 16 patients on ICI therapy. The Wilcoxon rank-sum test was used to determine the significant change in the initial and final ocular surface indices.

RESULTS

Fifty percent of the eyes demonstrated worsening Schirmer I scores; 29% showed an increase in lissamine green staining. During follow-up, 43% of patients experienced a decline in OSDI scores. Treatments included preservative-free artificial tears (88%), cyclosporine (25%), topical corticosteroids (31%), warm compresses (25%); punctal plugs (13%). Median follow-up time was 3.4 months (range:0-79 ); median ICI treatment duration was 7 months (range:1-40). Four patients died during the observation period.

CONCLUSION

A significant proportion of patients experience changes in ocular surface markers while treated with ICIs. Medical intervention can lead to stabilization of ocular surface disease.

Effectiveness of difluprednate in addition to systemic therapy for the treatment of anterior scleritis

AIM

To describe the effectiveness and side-effect profile of adding difluprednate therapy to patients with anterior scleritis being treated systemically.

METHODS

Retrospective chart review. Charts from all patients with anterior scleritis who were treated with topical difluprednate in addition to systemic therapy from 1 January 2018 to 1 January 2020 were reviewed. Data collected included: demographics, scleritis type, systemic diagnosis, presence of nodules or necrosis, changes in scleritis activity, intraocular pressure (IOP), number of difluprednate drops used, type of systemic treatment used, best-corrected visual acuity (BCVA) and lens status. The primary outcome was clinical resolution of scleritis. Secondary outcomes included BCVA loss ≥2 lines, change in lens status or cataract surgery and IOP ≥24 mm Hg.

RESULTS

Thirty-two patients (44 eyes) were analysed. The median age was 57 years (IQR 52, 72); 59% were female; 72% were Caucasian. An associated systemic disease was present in 59%. Systemic therapies used when difluprednate was added were: 65% immunosuppressive agents, 43% prednisone and 25% non-steroidal anti-inflammatory drugs. The addition of difluprednate resulted in clinical resolution in 79.6% of the treated eyes. Median time to inactivity was 9 weeks (IQR 5, 20). Eyes initially using 2-4 drops per day had a higher response rate (89%, p=0.005). Over a median follow-up of 34 weeks (IQR 21, 74), 11 eyes had IOP elevation; 6 eyes lost ≥2 lines of BCVA, 5 eyes had cataract progression.

CONCLUSION

Most eyes treated with difluprednate achieved inactivity. The addition of difluprednate to systemic therapies provides an alternative to achieve control of inflammation.

Analyzing the demographics of patients with uveitis in an indigent, urban population

PURPOSE

To study the types of uveitis examined in a hospital serving indigent populations in need of low-cost care.

METHODS

A retrospective chart review examined the electronic medical records of all patients with uveitis-related at Drexel Eye Physicians. Data collected included demographics, anatomic location of the uveitis, systemic disease associations, treatment modalities and insurance. Statistical analysis was performed using χ² or Fischer exact tests.

RESULTS

270 patients (366 eyes) were included for analysis, 67% of patients identified as African American. Most eyes (95.3%, N = 349) were treated with topical corticosteroid drops, and only 6 (1.6%) received an intravitreal implant. Immunosuppressive medications were started in 24 patients (8.9%). Nearly 80% depended to some extent on Medicare or Medicaid Assistance for treatment coverage. There was no association between insurance type and use of biologics or difluprednate.

CONCLUSION

We found no association between insurance type and the prescription of medications for uveitis that should be used at home. There was a minimal number of patients prescribed medications for implantation in the office. The adherence of use of medications at home should be investigated.

Effectiveness of Difluprednate for the Treatment of Anterior Scleritis

PURPOSE

To describe the effectiveness and side effect profile of difluprednate therapy in a series of patients with anterior scleritis.

DESIGN

Retrospective, interventional case series.

METHODS

Data collected from all patients with anterior scleritis who used difluprednate as a single treatment agent from January 1, 2018, to January 1, 2020, including demographics, scleritis type, presence of nodules or necrosis, changes in scleritis activity, intraocular pressure (IOP), number of difluprednate drops used, best-corrected visual acuity (BCVA), and lens status. The primary outcome was clinical resolution of scleritis. Secondary outcomes included BCVA loss ≥2 lines, change in lens status or cataract surgery, and IOP ≥24 mm Hg.

RESULTS

Twenty-five patients (35 eyes) were analyzed. The median age was 60 years (range 13-78); 60% were female; 64% were White. Forty percent had bilateral disease, and 44% of patients had an associated systemic disease. The majority of eyes (66%) had diffuse anterior scleritis. Eighty-three percent of eyes achieved resolution of scleritis, with a median time of resolution of 6 weeks. Eyes treated with an initial dose of ≥4 times daily were more likely to achieve disease resolution (hazard ratio [HR] = 3.43, 95% confidence interval [CI] 1.19, 9.88, P = .02). Nine eyes had IOP elevation. Four eyes lost ≥2 lines of BCVA, and 1 due to cataract progression. One eye underwent cataract surgery.

CONCLUSIONS

Difluprednate alone may effectively treat non-infectious anterior scleritis with a tolerable side effect profile.

Clinical Outcomes in Vitrectomized versus Non-vitrectomized Eyes in Patients with Primary Vitreoretinal Lymphoma

AIM

To evaluate if there is a difference in the clinical course of primary vitreoretinal lymphoma (PVRL) in vitrectomized versus non-vitrectomized eyes.

METHODS

Observational multicenter retrospective case series of patients diagnosed with PVRL between 2007 and 2019, at three tertiary centers. The main outcomes were relapse rates, inflammatory parameters, and best-corrected visual acuities (BCVA). Statistical methods used were an adjusted generalized estimating equation model, and a proportional Cox model.

RESULTS

Eighty patients (150 eyes) were followed for a median of 1.7 years. At presentation, there were no clinical differences between the groups. The relapse rate was 0.091/eye-year (EY) for vitrectomized eyes and 0.087/EY for non-vitrectomized eyes (p = .35). Vitrectomized eyes had better BCVA than non-vitrectomized eyes (p < .001).

CONCLUSIONS

Vitrectomy had no effect on the relapse rate in eyes with PVRL. However, the decrease in vitreous cell and debris led to vitrectomized eyes having better visual acuity than non-vitrectomized eyes.

Attitudes toward parental leave and breastfeeding during ophthalmology residency

INTRODUCTION

As mandated by the Accreditation Council for Graduate Medical Education (ACGME), residency programs are required to have parental leave policies. A lack of standardized requirements leads to a lack of uniformity among programs. We discuss resident and program director attitudes toward parental leave and examine the range of policies on parental leave and breastfeeding within ophthalmology residency programs.

METHODS

Two electronic surveys assessing perceptions toward parental leave during residency and breastfeeding on return to clinical duties were created individually for completion by ophthalmology residents or residency program directors, respectively, with responses collected over 4 weeks.

RESULTS

Of residents who took parental leave, 23 (87%) denied taking time off without pay. The most commonly reported effects on training by residents were missed surgical training and impact on research. Nearly 60% of residents (N = 26) reported receiving negative feedback or actions prior to or after the leave. The majority of residents felt program directors and coresidents were supportive (53.8%, 48.1%, respectively), but parental leave negatively affected their coresidents (46.2%). Twenty-five program directors reported that there are written parental leave policies in place at their institution. Sex disparities were noted, with program directors reporting more negative impacts on surgical training in female residents (p = 0.035). There was no statistically significant difference between program director attitudes on clinical training, well-being, or burnout by resident sex. All program directors were supportive of breastfeeding; half reported an institutional breastfeeding policy.

CONCLUSION

A national discussion on standardizing parental leave and breastfeeding policies over all ophthalmology residency programs is warranted.

Changes in patient visits and diagnoses in a large academic center during the COVID-19 pandemic

Background

To minimize the risk of viral transmission, ophthalmology practices limited face-to-face encounters to only patients with urgent and emergent ophthalmic conditions in the weeks after the start of the COVID-19 epidemic in the United States. The impact of this is unknown.

Methods

We did a retrospective analysis of the change in the frequency of ICD-10 code use and patient volumes in the 6 weeks before and after the changes in clinical practice associated with COVID-19.

Results

The total number of encounters decreased four-fold after the implementation of clinic changes associated with COVID-19. The low vision, pediatric ophthalmology, general ophthalmology, and cornea divisions had the largest total decrease of in-person visits. Conversely, the number of telemedicine visits increased sixty-fold. The number of diagnostic codes associated with ocular malignancies, most ocular inflammatory disorders, and retinal conditions requiring intravitreal injections increased. ICD-10 codes associated with ocular screening exams for systemic disorders decreased during the weeks post COVID-19.

Conclusion

Ophthalmology practices need to be prepared to experience changes in practice patterns, implementation of telemedicine, and decreased patient volumes during a pandemic. Knowing the changes specific to each subspecialty clinic is vital to redistribute available resources correctly.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12886-021-01886-7.

Adalimumab-associated Acquired Hemophilia in a Patient with Scleritis

Acquired hemophilia A (AHA) is a rare condition that may be drug-induced. In this case report, we describe a patient who presented with extensive subcutaneous bleeding three years after beginning treatment with adalimumab for necrotizing scleritis. His workup was compatible with drug-induced AHA. He was treated with high-dose corticosteroids, cyclophosphamide, and rituximab. Adalimumab was discontinued. We present this case as an example of a rare, but potentially life-threatening, complication of adalimumab.

Escalation to Weekly Adalimumab for the Treatment of Ocular Inflammation

Purpose: To report the outcomes of the escalation of adalimumab (ADA) dose for refractory ocular inflammatory diseases.Methods: A retrospective case series of 15 patients (29 eyes) diagnosed with ocular inflammatory disease, including uveitis and scleritis, which was not adequately controlled with standard, every other week ADA dosing, leading to an escalation to weekly dosing.Results: Ten of fifteen patients escalated to weekly ADA achieved control of their inflammation; neither of the two patients increased for control of cystoid macular edema (CME) had resolution and required regional corticosteroids. One patient discontinued weekly ADA due to serious infection. The median length of follow up was 12 months.Conclusion: Our series suggests that the escalation of ADA can be a useful strategy for treating recalcitrant ocular inflammation, but may not be adequate to treat refractory CME.

Outcome of phacoemulsification and intraocular lens implantion in eyes with pseudoexfoliation and weak zonules

PURPOSE

To evaluate the results of phacoemulsification and intraocular lens (IOL) implantation in eyes with pseudoexfoliation (PEX) and zonular weakness.

METHODS

Two hundred and ninety-five patients (295 eyes) were recruited. This study is a retrospective case review of 143 PEX eyes with preoperative (Group 1) and intraoperative (Group 2) signs of zonular weakness, 76 PEX eyes (Group 3) and 76 non-PEX eyes without zonular weakness (Group 4) undergoing phacoemulsification and IOL implantation performed by single surgeon from year 1988 to 2010. Outcome measures included frequency of postoperative pseudophakodonesis, anterior capsule contraction and IOL decentration/dislocation. Postoperative vision, intraocular pressure (IOP) and glaucoma medication requirements were also assessed. A subanalysis on identical postoperative outcomes of capsular tension ring (CTR) implantation was performed. Comparisons were made for 52 PEX eyes with zonular weakness receiving nonsutured CTR implantation (Group A), 91 PEX eyes with zonular weakness not receiving CTR implantation (Group B) and 142 PEX eyes without zonular weakness (Group C).

RESULTS

Mean follow-up was 4.08 ± 0.48 years and maximum was 8.3 years. Both Groups 1 and 2 showed higher incidence of complications than Groups 3 and 4 (p < 0.01). Vision, IOP and glaucoma medication requirements were similar postoperatively for all four groups. Groups A and B had a significantly higher proportion of postoperative complications than Group C (p < 0.01). There was no significant difference in risk of developing these complications between Groups A and B (p = 0.64).

CONCLUSION

Postoperative complication rates are higher in PEX eyes with preoperative and/or intraoperative signs of zonular weakness undergoing phacoemulsification. Nonsutured CTR implantation does not prevent IOL and capsular complications postoperatively.

Clinical Features and Prognostic Factors in Presumed Ocular Tuberculosis

PURPOSE

To characterize the clinical features in patients with presumed ocular tuberculosis (TB) and determine prognostic factors of visual outcomes and complications in this disease.

MATERIAL AND METHODS

Retrospective case series of 35 patients (29 females, 6 males) with presumed ocular TB from referral centers in Chile and Spain between 2002 and 2012. Medical records were reviewed, and data regarding clinical features, complications, best-corrected visual acuity (BCVA), duration of disease, extraocular manifestations, and therapy were retrieved. Prognostic factors for low vision (BCVA 20/50 or less), legal blindness (BCVA 20/200 or less), and complications (cataract, glaucoma, and macular lesion) were evaluated. To calculate correlations, we used Spearman's rank correlation test. To determine clinical predictors, we used the binary logistic regression test.

RESULTS

Anterior and non-granulomatous uveitis was the most common types of inflammation. Only 2 (5.7%) patients had respiratory symptoms, and 6 (17.1%) patients had an abnormal chest X-ray at diagnosis. All patients received combined antitubercular therapy with a mean duration of 6.9 ± 2.3 months. A longer duration of symptoms at diagnosis was associated with both low vision and legal blindness. Older patients had a higher risk of legal blindness. A longer duration of symptoms as well as anterior inflammation demonstrated an increased risk for cataract formation. The duration of the symptoms and baseline BCVA had a positive correlation with the final BCVA. Prognostic factors of macular lesions were not found.

CONCLUSIONS

The diagnosis of ocular TB can be difficult due to the lack of extraocular manifestations and the broad spectrum of ocular features. A longer duration of symptoms at diagnosis was associated with poorer visual outcomes and cataracts. Therefore, efforts should be made to avoid a delay in the diagnosis of ocular TB and to identify prognostic factors for visual outcomes and complications.

Evaluation of the Accuracy of T-SPOT.TB for the Diagnosis of Ocular Tuberculosis in a BCG-vaccinated, Non-endemic Population

PURPOSE

To determine the performance of T-SPOT.TB, an interferon gamma release assay test, in patients with ocular tuberculosis (TB) in a BCG-vaccinated, non-endemic population.

METHODS

We employed a nested case-control design. In total, 45 subjects were enrolled (23 patients with ocular tuberculosis and 22 patients with other causes of uveitis). A blood sample was collected from each subject, and T-SPOT.TB was executed. Laboratory professionals were blinded to the disease status of each subject.

RESULTS

Five patients were excluded because of indeterminate results. The calculated sensitivity and specificity were 0.80 and 0.85, respectively. The positive likelihood ratio was 5.33 and the negative likelihood ratio was 0.23. The overall accuracy of the test was 0.83.

CONCLUSIONS

T-SPOT.TB adequately diagnosed ocular TB. This technique is particularly useful in populations where BCG vaccinations are still mandatory.