Colorectal cancer (CRC) can be effectively prevented or detected with guideline concordant screening, yet Medicaid enrollees experience disparities. We used microsimulation to project CRC screening patterns, CRC cases averted, and life-years gained in the population of 68,077 Oregon Medicaid enrollees 50-64 over a five year period starting in January 2019. The simulation estimated the cost-effectiveness of five intervention scenarios - academic detailing plus provider audit and feedback (Detailing+), patient reminders (Reminders), mailing a Fecal Immunochemical Test (FIT) directly to the patient's home (Mailed FIT), patient navigation (Navigation), and mailed FIT with Navigation (Mailed FIT + Navigation) - compared to usual care. Each intervention scenario raised CRC screening rates compared to usual care, with improvements as high as 11.6 percentage points (Mailed FIT + Navigation) and as low as 2.5 percentage points (Reminders) after one year. Compared to usual care, Mailed FIT + Navigation would raise CRC screening rates 20.2 percentage points after five years - averting nearly 77 cancer cases (a reduction of 113 per 100,000) and exceeding national screening targets. Over a five year period, Reminders, Mailed FIT and Mailed FIT + Navigation were expected to be cost effective if stakeholders were willing to pay $230 or less per additional year up-to-date (at a cost of $22, $59, and $227 respectively), whereas Detailing+ and Navigation were more costly for the same benefits. To approach national CRC screening targets, health system stakeholders are encouraged to implement Mailed FIT with or without Navigation and Reminders.
Medical image classification often relies on CNNs to capture local details (e.g., lesions, nodules) or on transformers to model long-range dependencies. However, each paradigm alone is limited in addressing both fine-grained structures and broader anatomical context. We propose ConvTransGFusion, a hybrid model that fuses ConvNeXt (for refined convolutional features) and Swin Transformer (for hierarchical global attention) using a learnable dual-attention gating mechanism. By aligning spatial dimensions, scaling each branch adaptively, and applying both channel and spatial attention, the proposed architecture bridges local and global representations, melding fine-grained lesion details with the broader anatomical context essential for accurate diagnosis. Tested on four diverse medical imaging datasets—including X-ray, ultrasound, and MRI scans—the proposed model consistently achieves superior accuracy, precision, recall, F1, and AUC over state-of-the-art CNNs and transformers. Our findings highlight the benefits of combining convolutional inductive biases and transformer-based global context in a single learnable framework, positioning ConvTransGFusion as a robust and versatile solution for real-world clinical applications.
Although screening is effective in reducing incidence, mortality, and costs of treating colorectal cancer (CRC), it remains underutilized, in part due to limited insurance access. We used microsimulation to estimate the health and financial effects of insurance expansion and reduction scenarios in North Carolina (NC). We simulated the full lifetime of a simulated population of 3,298,265 residents age-eligible for CRC screening (ages 50-75) during a 5-year period starting January 1, 2018, including polyp incidence and progression and CRC screening, diagnosis, treatment, and mortality. Insurance scenarios included: status quo, which in NC includes access to the Health Insurance Exchange (HIE) under the Affordable Care Act (ACA); no ACA; NC Medicaid expansion, and Medicare-for-all. The insurance expansion scenarios would increase percent up-to-date with screening by 0.3 and 7.1 percentage points for Medicaid expansion and Medicare-for-all, respectively, while insurance reduction would reduce percent up-to-date by 1.1 percentage points, compared to the status quo (51.7% up-to-date), at the end of the 5-year period. Throughout these individuals' lifetimes, this change in CRC screening/testing results in an estimated 498 CRC cases averted with Medicaid expansion and 6031 averted with Medicare-for-all, and an additional 1782 cases if health insurance gains associated with ACA are lost. Estimated cost savings - balancing increased CRC screening/testing costs against decreased cancer treatment costs - are approximately $30 M and $970 M for Medicaid expansion and Medicare-for-all scenarios, respectively, compared to status quo. Insurance expansion is likely to improve CRC screening both overall and in underserved populations while saving money, with the largest savings realized by Medicare.
Of the nearly four million births that occur each year in the United States, almost one in three is a cesarean delivery. Despite the increasing C-section rate over the years, there is no evidence that the increase has caused a decrease in neonatal or maternal mortality or morbidity. Bayesian decision analysis is used to model the decision between classifying a patient as “failure-to-progress,” which is cause for a C-section, using current information (prior probability) or information gathered (posterior probability) as labor continues. The Bayesian decision models determine the conditions under which it is appropriate to gather additional information (i.e., take an observation) before deciding to end labor and perform a C-section based on the decision maker’s belief about successful labor. During an observation period, the decision maker learns more about the patient and her medical state and the likelihood of a successful vaginal delivery is updated. This study determines the conditional value of information (conditional on the decision maker’s prior belief) and determines the conditions under which information has positive value. This model can be used to facilitate shared decision making for labor and delivery through communicating beliefs, risk perceptions, and the associated actions. The online appendix is available at https://doi.org/10.1287/serv.2018.0205 .