Predicting ophthalmic clinic non-attendance using machine learning: Development and validation of models using nationwide data.

BACKGROUND: Ophthalmic clinic non-attendance in New Zealand is associated with poorer health outcomes, marked inequities and costs NZD$30 million per annum. Initiatives to improve attendance typically involve expensive and ineffective brute-force strategies. The aim was to develop machine learning models to accurately predict ophthalmic clinic non-attendance. METHODS: This multicentre, retrospective observational study developed and validated predictive models of clinic non-attendance. Attendance data for 3.1 million appointments from all New Zealand government-funded ophthalmology clinics from 2009 to 2018 were aggregated for analysis. Repeated ten-fold cross validation was used to train and optimise XGBoost and logistic regression models on several demographic and clinic-related variables. Models developed using the entire training set were compared with those restricted to regional subsets of the data. RESULTS: In the testing data set from 2019, there were 407 574 appointments (median [range] age, 66 [0-105] years; 210 365 [51.6%] female) with a non-attendance rate of 5.7% (n = 23 309 missed appointments), XGBoost models trained on each region's data achieved the highest mean AUROC of 0.764 (SD 0.058) and mean AUPRC of 0.157 (SD 0.072). XGBoost performed better than logistic regression (mean AUROC = 0.756, p = 0.002). Training individual XGBoost models for each region led to better performance than training a single model on the complete nationwide dataset (mean AUROC = 0.754, p = 0.04). CONCLUSION: Machine learning algorithms can predict ophthalmic clinic non-attendance with relatively basic demographic and clinic data. These findings suggest further research examining implementation of such algorithms in scheduling systems or public health interventions may be useful.

Ophthalmic surgery in New Zealand: analysis of 410,099 surgical procedures and nationwide surgical intervention rates from 2009 to 2018.

BACKGROUND: Surgical intervention rates (SIR) provide a proxy measure of disease burden, surgical capacity, and the relative risk-benefit ratio of surgery. The current study assessed decade trends in ophthalmic surgery and calculated SIRs for all major classes of commonly performed ophthalmic procedures in New Zealand. METHODS: Retrospective population-based analysis of all ophthalmic surgical procedures performed in New Zealand from 2009 to 2018. National and regional datasets from public and private health sectors and industry were analysed. SIRs were calculated for all major ophthalmic procedures, and subgrouped by patient demographics. RESULTS: There were 410,099 ophthalmic surgical procedures completed with a 25.3% overall increase over 10 years. Procedures were mostly government-funded (51%, n = 210,830) with 71% of patients aged over 64 years. Cataract surgery (78%, n = 318,564) had the highest mean SIR (703/100,000/year) and increased by 25% during the study period, consistent with population growth in the over 64 years old age group. Vitrectomy surgery had the second highest mean SIR (67/100,000/year) and increased by 50%, well above national population growth during the study period. Other SIRs included conjunctival lesion-biopsy (38/100,000/year), glaucoma (33/100,000/year), strabismus (20/100,000/year), dacryocystorhinostomy (10/100,000/year), and keratoplasty surgery (4/100,000/year). CONCLUSIONS: This comprehensive review of New Zealand ophthalmic surgery reports increasing SIRs that cannot be explained by population growth alone. Cataract surgery numbers increased year on year consistent with the increase in the over 64 years old population. Vitrectomy surgery growth exceeded that of the national population, including those over 64 years.

Ocular Surface Squamous Neoplasia: A 12-Month Prospective Evaluation of Incidence in Waikato, New Zealand.

Ocular surface squamous neoplasia (OSSN) has a high incidence in the southern hemisphere. This prospective study evaluated the incidence of OSSN in the Waikato region of New Zealand. All patients presenting with pterygium or conjunctival lesions in the Waikato region in 2020 were included. All surgeons in the region were asked to send all conjunctival and corneal specimens excised for histopathologic examination. The primary outcome measure was the incidence of OSSN. Eighty-eight percent of all excised specimens were sent for histopathologic examination. Of the 185 excised lesions sent for histopathological assessment, 18 (10%) were reported as OSSN. Patients were on average 69.4 years of age (standard deviation, SD = 6.9), predominantly male (78%), and of New Zealand-European ethnicity (89%). The OSSN annual incidence was 3.67/100,000/year. Histology grades included conjunctival intraepithelial neoplasia (CIN)-I (25%), CIN-II (25%), CIN-III (12.5%), carcinoma in situ (25%), and invasive squamous cell carcinoma (SCC) (12.5%). One patient with invasive SCC required exenteration. This study identified a high incidence rate of OSSN and is the first prospective study to analyze OSSN epidemiology in New Zealand.

Ocular surface squamous neoplasia in New Zealand: a ten-year review of incidence in the Waikato region.

BACKGROUND: Ocular surface squamous neoplasia (OSSN) is a relatively rare disease with a low mortality and highly variable incidence. Despite a high incidence of OSSN in the Southern hemisphere, there is limited epidemiological data for New Zealand. The current study aims to assess the incidence, demographics, and histological grade of OSSN in the Waikato region of New Zealand, home to ~10% of the population of New Zealand. METHODS: Non-interventional retrospective cohort study. All conjunctival biopsy histology reports from 2010 to 2019 in the Waikato region of New Zealand were analysed. Age, sex, and ethnicity were analysed and the incidence of OSSN was calculated. Main outcome measures included histological grade, rate of recurrence, and incidence of OSSN. RESULTS: A total of 386 patients underwent conjunctival biopsy with histology during the study period. Eighty-three lesions (22%) involving 80 patients (21%) were reported positive for OSSN. Patients with OSSN had a mean age of 68.9 years (SD = 13.2), were predominantly male (76%), and of New Zealand-European ethnicity (53%). Conjunctival intraepithelial neoplasia-1 (30%) was the most frequent diagnosis. Three patients (4%) had recurrent lesions requiring repeat biopsy. The peak annual OSSN incidence rate was 3.81/100,000 population in 2019. The overall ten-year incidence was 2.13/100,000 population/year. CONCLUSION: This is the largest study to investigate OSSN incidence in New Zealand. The incidence rate of OSSN is one of the highest rates reported in the literature.

Rates of spondyloarthropathies vary with age and ethnicity in HLAB27 uveitis.

BACKGROUND/AIMS: To determine associations between HLAB27-positive uveitis, ethnicity and seronegative spondyloarthropathies (SpAs) in a New Zealand population. METHODS: Retrospective observational cohort study. Medical records of all subjects with uveitis at Auckland District Health Board from 2008 to 2018 were reviewed for HLAB27 status, age of presentation, ethnicity and SpA. RESULTS: In 10 years, 2567 subjects with uveitis were seen and 492 (19.2%) were HLAB27-positive. Of the HLAB27-positive subjects, 301 were male (60.3%) and median age was 37.8 years (IQR 29.7-50.0). Ethnicities were Caucasian (n=298, 60.6%), Asian (n=111, 22.6%), Maori (n=41, 8.2%) and Pacific Islander (n=38, 7.7%). Uveitis classification was anterior (n=478, 97.2%), intermediate (n=40, 8.1%), panuveitis (n=9, 1.8%) and scleritis (n=2, 0.4%). Maori or Pacific Islander ethnicity was associated with intermediate or panuveitis (p=0.003). Ankylosing spondylitis occurred in 163 subjects (33.1%); 29 (17.8%) were Maori or Pacific Islander. Subjects were younger (OR 0.982, p=0.009) and male (OR 1.980, p=0.001). There was no association with ethnicity or uveitis classification. Psoriatic arthritis (PsA) occurred in 11 subjects (2.2%). Chronic anterior uveitis was more common with PsA (27.3% vs 7.1%, p=0.023). There was no association with gender or ethnicity. Inflammatory bowel disease occurred in 19 subjects (3.8%) and reactive arthritis occurred in 14 subjects (2.8%). None developed chronic anterior uveitis (p=0.246 and p=0.227, respectively). There was no association with age at presentation, gender, ethnicity or uveitis classification. CONCLUSION: This cohort of New Zealand-based subjects with HLAB27-positive uveitis showed a difference in age and ethnicity in uveitis subtypes and SpAs.

Inflammatory-induced astigmatism: acute changes in corneal curvature secondary to marginal keratitis and previous mitomycin-C treatment.

A 52-year-old man presented with left eye redness, blurred vision, and photophobia. A history included marginal keratitis and conjunctival squamous cell carcinoma treated with excision and topical mitomycin-C. Examination revealed current marginal keratitis, managed with topical antibiotic and corticosteroid. Regular assessments included high-resolution optical coherence corneal tomography, refraction, and blood tests to exclude other causes of peripheral infiltrate and thinning. Two weeks later, visual acuity (VA) decreased and astigmatism increased. Significant refractive instability with astigmatism increased to 5.25 diopters with a corresponding decrease in VA. After 4 months of topical treatment, the marginal keratitis, astigmatism, and change in VA resolved. To the authors' knowledge, this is the first case report to describe an induced and fluctuating high-magnitude corneal astigmatic change in response to marginal keratitis.

East London deliberate corrosive fluid injuries.

OBJECTIVES: The aims of this case series are to (1) highlight the incidence of deliberate corrosive fluid injuries (DCFI) in East London; (2) ascertain the types of substances used; and (3) grade the resulting ocular surface burns using two validated grading systems. METHODS: The Metropolitan Police 2017 Freedom of Information (FOI) requests were reviewed for DCFI. We retrospectively reviewed patient records between October 2016 and 2017 from local A&E departments for DCFI. Roper-Hall and Dua Classifications were used to classify prognoses in patients with deliberate ocular corrosive fluid injuries (DOCFI). RESULTS: The FOIs demonstrated an increasing trend with the highest number of DCFI surrounding our centre. We identified 57 patients with DCFI. Based on pre-irrigation pH measurement, 54.4% (n = 31) sustained acidic; 21.1% (n = 12) alkaline and 24.6% (n = 14) unknown injuries. More than half, 66.7% (n = 38), sustained DOCFI with median presentation visual acuity (VA) 0.25 LogMAR (n = 26, IQR: 0.10-0.39) and 1 month VA 0.20 LogMAR (n = 16, IQR: 0.02-0.20). Prognosis of the worse eye using (1) Roper-Hall Classification was 55% good (Grade I-II); 23% guarded (Grade III); 23% guarded (Grade IV); and (2) Dua Classification was 23% very good (Grade I); 35% good (Grade II-III); 19% good to guarded (Grade IV); 23% guarded to poor (Grade V); and 0% very poor (Grade VI). CONCLUSIONS: DCFI mostly affected young men in this cohort. Our clinical experience suggested an alarming increase in DCFI and is supported by police data. Although this patient group appears to show relatively good outcomes, legislation to prevent these attacks should be prioritised.

Genetic variation of genes involved in dihydrotestosterone metabolism and the risk of prostate cancer.

PURPOSE: Dihydrotestosterone (DHT) is an important factor in prostate cancer (PCA) genesis and disease progression. Given PCA's strong genetic component, we evaluated the possibility that variation in genes involved in DHT metabolism influence PCA risk. EXPERIMENTAL DESIGN: We investigated copy number variants (CNV) and single nucleotide polymorphisms (SNP). We explored associations between CNV of uridine diphospho-glucuronosyltransferase (UGT) genes from the 2B subclass, given their prostate specificity and/or involvement in steroid metabolism and PCA risk. We also investigated associations between SNPs in genes (HSD3B1, SRD5A1/2, and AKR1C2) involved in the conversion of testosterone to DHT, and in DHT metabolism and PCA risk. The population consisted of 426 men (205 controls and 221 cases) who underwent prostate-specific antigen screening as part of a PCA early detection program in Tyrol, Austria. RESULTS: No association between CNV in UGT2B17 and UGT2B28 and PCA risk was identified. Men carrying the AA genotype at SNP rs6428830 (HSD3B1) had an odds ratio (OR) of 2.0 [95% confidence intervals (95% CI), 1.1-4.1] compared with men with GG, and men with AG or GG versus AA in rs1691053 (SRD5A1) had an OR of 1.8 (95% CI, 1.04-3.13). Individuals carrying both risk alleles had an OR of 3.1 (95% CI, 1.4-6.7) when compared with men carrying neither (P = 0.005). Controls with the AA genotype on rs7594951 (SRD5A2) tended toward higher serum DHT levels (P = 0.03). CONCLUSIONS: This is the first study to implicate the 5alpha-reductase isoform 1 (SRD5A1) and PCA risk, supporting the rationale of blocking enzymatic activity of both isoforms of 5alpha-reductase for PCA chemoprevention.

ERG rearrangement metastasis patterns in locally advanced prostate cancer.

OBJECTIVES: To interrogate multifocal prostate cancer (PCa) to determine its predilection for metastasis, using ERG rearrangement as marker of clonality. A hallmark of PCa is that distinct tumor foci may arise independently, which has important biological and clinical implications. Recent studies characterizing ERG-rearranged PCa possessing intrafocal homogeneity but interfocal heterogeneity support this hypothesis. METHODS: We studied 26 patients who underwent prostatectomy and lymphadenectomy with at least 2 distinct PCa foci and 1 lymph node (LN) metastasis. Each focus was assessed for size, Gleason score, ERG rearrangement, and TMPRSS2-ERG transcript. RESULTS: Fifteen of 26 cases exhibited interfocal homogeneity with regard to ERG rearrangement (ie, presence vs absence of ERG rearrangement). ERG rearrangement was present in all foci for 6 and absent in all foci for 9 cases. Two cases revealed interfocal heterogeneity with regard to rearrangement mechanism (ie, rearrangement through insertion or deletion). Eight of 26 cases revealed interfocal heterogeneity with regard to rearrangement status. In all cases with at least 1 ERG rearranged focus, we found the corresponding LN metastasis harboring an ERG rearrangement. Interestingly, in a subset of cases the rearrangement status in the LN did not correspond to size or Gleason score. All but 2 ERG rearranged foci had detectable TMPRSS2-ERG transcript levels. CONCLUSIONS: When multifocal PCa demonstrates both ERG-positive and ERG-negative foci, the positive foci have a greater predilection for metastasis. Larger studies are needed to confirm the potential additional risk an ERG rearranged focus confers on the likelihood of disease progression.