A361with the observed values. Results: The squared errors for the NB distributions were consistently lower than the Poisson and binomial fits. Across the full 64 weeks, the total errors (average errors) were 0.1506 (0.009), 0.5493 (0.034) and 0.7003 (0.044) for the NB, Poisson and binomial distributions, respectively. NB distributions also consistently provided a better visual fit to the observed values. ConClusions: The use of NB distributions with estimated dispersion parameters provided estimates with lower errors in comparison with Poisson or binomial fits to the distribution of patients by headache day frequency. This approach provides a solution where distribution parameters cannot be observed directly, and facilitates the modeling of comparisons for which patient level data are not available.

No abstract

A733to show that raw and aggregated data can be transmitted remotely outside of the traditional clinical setting. This research demonstrates that data from accelerometers has value beyond the traditional sleep and activity endpoints and could be used in remote studies. Methods: Two healthy volunteers (A and B) were provided with accelerometers and hubs. The hubs were SIM enabled to allow for continuous data collection. Volunteer A wore the device for 24 hours and used a diary to identify 27 scratching events of approximately 30 seconds duration. Volunteer B wore the device for 8 hours and used a diary to identify 7 scratching events. Results: Raw 100 Hz accelerometer data was transmitted remotely via the hubs to the centralized study center, from where it was further processed and analyzed. An analytical model was developed using the data from Volunteer A to identify scratching events at a 10 second epoch level. This algorithm achieved sensitivity and specificity values of 99 and 100%, respectively for Volunteer A. The algorithm was further evaluated on unseen (from the model's point of view) Volunteer B and achieved sensitivity and specificity values of 99 and 86%, respectively. ConClusions: Accelerometerbased wearables are gaining acceptance in clinical trials as a means of generating objective endpoints for sleep and activity using validated algorithms. This study has shown that the application of suitable algorithms to raw accelerometer data has the potential to generate clinically relevant outcome measures associated with patient motor movement patterns, which can have significance in studies looking at tremor and itch and other clinical symptoms. The ability to generate and transmit raw data from a patient's home facilitates the integration of this methodology into remote and virtual trials.

defined by clinical experts considering classical therapy and gene therapy with three levels of efficacy, tolerability, and price (low, medium and high) from first line to third line. The possibility of a one-time cost was implemented and the model was updated with input data from the literature for the UK settings. Model outputs include time in different health states, quality adjusted life years (QALYs) and costs from National Health Service and societal perspectives. The model remains open source and is now available on GitHub. RESULTS: The largest cost components from the payer perspective were the price of the new drug followed by physician visits and hospitalisations. The predicted costs and QALYs were driven by utility values, effectiveness, and frequency of physician visits. Differences in QALYs and costs between strategies were dependent with the duration of the simulation: a shorter time horizon limited the capture of the long term health benefit, increasing the impact of the high one time-cost on the ICER. CONCLUSIONS: The discrete event simulation model can provide a comprehensive evaluation of different therapeutic options in MDD including potential gene therapy and can be used to compare a wide range of health care technologies in various groups of patients in MDD. The fact that the DES is open-source reinforced the credibility and the acceptability of the model.

S9 ObjectiveS: To assess the cost-effectiveness of fibrosis-based direct-acting antiviral (DAA) treatment policies for chronic Hepatitis C (HCV) patients at the Kaiser Permanente Mid-Atlantic States health system (KPMAS). MethOdS: We used a Markov model to compare the lifetime costs and effects of treating chronic HCV patients at different stages of disease severity based on a fibrosis score. The initial distribution of patients across fibrosis scores, the effectiveness of DAA therapy and follow-up and monitoring protocols used to build the model were specific to the KPMAS system. Other parameters, including direct and indirect costs, transition probabilities and health state utilities were derived from the literature. We performed both deterministic and probabilistic sensitivity analyses to assess the robustness of our results. ReSultS: The universal, or 'Treat All', treatment option was the dominant strategy from both the societal and health care sector perspectives. Varying the model parameters in a deterministic analysis did not change this conclusion. It is important to note that the range of incremental costs between the three less restrictive policies was very small-the difference between the 'Treat F1+' and the 'Treat All' option was just under $100 per person. Probabilistic sensitivity analyses showed, at both the $100,000/QALY and $150,000/QALY thresholds, there was a 70% chance that the 'Treat All' option was cost-effective and a 30% chance the 'Treat F1+' option was cost-effective. cOncluSiOnS: Our results are consistent with the current literature on the value of treating patients with the new DAA therapies-expanded treatment access is cost-effective and in many cases cost saving. While our results are primarily applicable to a unique integrated health care system, the results offer some direction to any health care setting that is faced with resource constraints in the face of these highly priced drugs.