Cost-effectiveness Analysis of a Personalized, Teleretinal-Inclusive Screening Policy for Diabetic Retinopathy via Markov Modeling.

PURPOSE: Although teleretinal imaging has proved effective in increasing population-level screening for diabetic retinopathy (DR), there is a lack of quantitative understanding of how to incorporate teleretinal imaging into existing screening guidelines. We develop a mathematical model to determine personalized DR screening recommendations that utilize teleretinal imaging and evaluate the cost-effectiveness of the personalized screening policy. DESIGN: A partially observable Markov decision process is employed to determine personalized screening recommendations based on patient compliance, willingness to pay, and A1C level. Deterministic sensitivity analysis was conducted to evaluate the impact of patient-specific factors on personalized screening policy. The cost-effectiveness of identified screening policies was evaluated via hidden-Markov chain Monte Carlo simulation on a data-based hypothetical cohort. PARTICIPANTS: Screening policies were simulated for a hypothetical cohort of 500 000 patients with parameters based on the literature and electronic medical records of 2457 patients who received teleretinal imaging from 2013 to 2020 from the Harris Health System. METHODS: Population-based mathematical modeling study. Interventions included dilated fundus examinations referred to as clinical screening, teleretinal imaging, and wait and watch recommendations. MAIN OUTCOME MEASURES: Personalized screening recommendations based on patient-specific factors. Accumulated quality-adjusted life-years (QALYs) and cost (USD) per patient under different screening policies. Incremental cost-effectiveness ratio to compare different policies. RESULTS: For the base cohort, on average, teleretinal imaging was recommended 86.7% of the time over each patient's lifetime. The model-based personalized policy dominated other standardized policies, generating more QALY gains and cost savings for at least 57% of the base cohort. Similar outcomes were observed in sensitivity analyses of the base cohort and the Harris Health-specific cohort and rural population scenario analysis. CONCLUSIONS: A mathematical model was developed as a decision support tool to identify a personalized screening policy that incorporates both teleretinal imaging and clinical screening and adapts to patient characteristics. Compared with current standardized policies, the model-based policy significantly reduces costs, whereas it is performing comparably, if not better, in terms of QALY gain. A personalized approach to DR screening has significant potential benefits that warrant further exploration. FINANCIAL DISCLOSURE(S): Proprietary or commercial disclosure may be found after the references.

State Variation in Squamous Cell Carcinoma of the Anus Incidence and Mortality, and Association With HIV/AIDS and Smoking in the United States.

PURPOSE: Squamous cell carcinoma of the anus (SCCA) incidence and mortality rates are rising in the United States. Understanding state-level incidence and mortality patterns and associations with smoking and AIDS prevalence (key risk factors) could help unravel disparities and provide etiologic clues. METHODS: Using the US Cancer Statistics and the National Center for Health Statistics data sets, we estimated state-level SCCA incidence and mortality rates. Rate ratios (RRs) were calculated to compare incidence and mortality in 2014-2018 versus 2001-2005. The correlations between SCCA incidence with current smoking (from the Behavioral Risk Factor Surveillance System) and AIDS (from the HIV Surveillance system) prevalence were evaluated using Spearman's rank correlation coefficient. RESULTS: Nationally, SCCA incidence and mortality rates (per 100,000) increased among men (incidence, 2.29-3.36, mortality, 0.46-0.74) and women (incidence, 3.88-6.30, mortality, 0.65-1.02) age ≥ 50 years, but decreased among men age < 50 years and were stable among similar-aged women. In state-level analysis, a marked increase in incidence (≥ 1.5-fold for men and ≥ two-fold for women) and mortality (≥ two-fold) for persons age ≥ 50 years was largely concentrated in the Midwestern and Southeastern states. State-level SCCA incidence rates in recent years (2014-2018) among men were correlated (r = 0.47, P < .001) with state-level AIDS prevalence patterns. For women, a correlation was observed between state-level SCCA incidence rates and smoking prevalence (r = 0.49, P < .001). CONCLUSION: During 2001-2005 to 2014-2018, SCCA incidence and mortality nearly doubled among men and women age ≥ 50 years living in Midwest and Southeast. State variation in AIDS and smoking patterns may explain variation in SCCA incidence. Improved and targeted prevention is needed to combat the rise in SCCA incidence and mitigate magnifying geographic disparities.

Cervical Cancer Incidence Among US Women, 2001-2019.

This study uses national cancer incidence data to evaluate calendar trends in cervical cancer incidence by age at diagnosis.

Kidney Level Lupus Nephritis Classification Using Uncertainty Guided Bayesian Convolutional Neural Networks.

The kidney biopsy based diagnosis of Lupus Nephritis (LN) is characterized by low inter-observer agreement, with misdiagnosis being associated with increased patient morbidity and mortality. Although various Computer Aided Diagnosis (CAD) systems have been developed for other nephrohistopathological applications, little has been done to accurately classify kidneys based on their kidney level Lupus Glomerulonephritis (LGN) scores. The successful implementation of CAD systems has also been hindered by the diagnosing physician's perceived classifier strengths and weaknesses, which has been shown to have a negative effect on patient outcomes. We propose an Uncertainty-Guided Bayesian Classification (UGBC) scheme that is designed to accurately classify control, class I/II, and class III/IV LGN (3 class) at both the glomerular-level classification task (26,634 segmented glomerulus images) and the kidney-level classification task (87 MRL/lpr mouse kidney sections). Data annotation was performed using a high throughput, bulk labeling scheme that is designed to take advantage of Deep Neural Network's (or DNNs) resistance to label noise. Our augmented UGBC scheme achieved a 94.5% weighted glomerular-level accuracy while achieving a weighted kidney-level accuracy of 96.6%, improving upon the standard Convolutional Neural Network (CNN) architecture by 11.8% and 3.5% respectively.