Both liver resection (LR) and cadaveric liver transplantation (CLT) are potentially curative treatments for patients with hepatocellular carcinoma (HCC) within the Milan criteria and with adequate liver function. Adopting either as a first‐line therapy carries major cost and resource implications. The objective of this study was to estimate the relative cost‐effectiveness of LR against CLT for patients with HCC within the Milan criteria using a decision analytic model. A Markov cohort model was developed to simulate a cohort of patients aged 55 years with HCC within the Milan criteria and Child‐Pugh A/B cirrhosis, undergoing LR or CLT, and followed up over their remaining life expectancy. Analysis was performed in different geographical cost settings: the USA, Switzerland and Singapore. Transition probabilities were obtained from systematic literature reviews, supplemented by databases from Singapore and the Organ Procurement and Transplantation Network (USA). Utility and cost data were obtained from open sources. LR produced 3.9 quality‐adjusted life years (QALYs) while CLT had an additional 1.4 QALYs. The incremental cost‐effectiveness ratio (ICER) of CLT versus LR ranged from $111,821/QALY in Singapore to $156,300/QALY in Switzerland, and was above thresholds for cost‐effectiveness in all three countries. Sensitivity analysis revealed that CLT‐related 5‐year cumulative survival, one‐time cost of CLT, and post‐LR 5‐year cumulative recurrence rates were the most sensitive parameters in all cost scenarios. ICERs were reduced below threshold when CLT‐related 5‐year cumulative survival exceeded 84.9% and 87.6% in Singapore and the USA, respectively. For Switzerland, the ICER remained above the cost‐effectiveness threshold regardless of the variations. Conclusion: In patients with HCC within the Milan criteria and Child‐Pugh A/B cirrhosis, LR is more cost‐effective than CLT across three different costing scenarios: the USA, Switzerland, Singapore. (Hepatology 2015;61:227–237)

Objectives. The objectives of this study are to develop a discrete-event simulation (DES) model for the Singapore Emergency Medical Services (EMS), and to demonstrate the utility of this DES model for the evaluation of different policy alternatives to improve ambulance response times. Methods. A DES model was developed based on retrospective emergency call data over a continuous 6-month period in Singapore. The main outcome measure is the distribution of response times. The secondary outcome measure is ambulance utilization levels based on unit hour utilization (UHU) ratios. The DES model was used to evaluate different policy options in order to improve the response times, while maintaining reasonable fleet utilization. Results. Three policy alternatives looking at the reallocation of ambulances, the addition of new ambulances, and alternative dispatch policies were evaluated. Modifications of dispatch policy combined with the reallocation of existing ambulances were able to achieve response time performance equivalent to that of adding 10 ambulances. The median (90th percentile) response time was 7.08 minutes (12.69 minutes). Overall, this combined strategy managed to narrow the gap between the ideal and existing response time distribution by 11–13%. Furthermore, the median UHU under this combined strategy was 0.324 with an interquartile range (IQR) of 0.047 versus a median utilization of 0.285 (IQR of 0.051) resulting from the introduction of additional ambulances. Conclusions. Response times were shown to be improved via a more effective reallocation of ambulances and dispatch policy. More importantly, the response time improvements were achieved without a reduction in the utilization levels and additional costs associated with the addition of ambulances. We demonstrated the effective use of DES as a versatile platform to model the dynamic system complexities of Singapore's national EMS systems for the evaluation of operational strategies to improve ambulance response times.

This paper describes the development of a deep learning model for prediction of hip fractures on pelvic radiographs (X-rays). Developed using over 40,000 pelvic radiographs from a single institution, the model demonstrated high sensitivity and specificity when applied to a test set of emergency department radiographs. This study approximates the real-world application of a deep learning fracture detection model by including radiographs with sub-optimal image quality, other non-hip fractures, and metallic implants, which were excluded from prior published work. The study also explores the effect of ethnicity on model performance, as well as the accuracy of visualization algorithm for fracture localization.