Machine learning for real-time aggregated prediction of hospital admission for emergency patients.

Machine learning for hospital operations is under-studied. We present a prediction pipeline that uses live electronic health-records for patients in a UK teaching hospital's emergency department (ED) to generate short-term, probabilistic forecasts of emergency admissions. A set of XGBoost classifiers applied to 109,465 ED visits yielded AUROCs from 0.82 to 0.90 depending on elapsed visit-time at the point of prediction. Patient-level probabilities of admission were aggregated to forecast the number of admissions among current ED patients and, incorporating patients yet to arrive, total emergency admissions within specified time-windows. The pipeline gave a mean absolute error (MAE) of 4.0 admissions (mean percentage error of 17%) versus 6.5 (32%) for a benchmark metric. Models developed with 104,504 later visits during the Covid-19 pandemic gave AUROCs of 0.68-0.90 and MAE of 4.2 (30%) versus a 4.9 (33%) benchmark. We discuss how we surmounted challenges of designing and implementing models for real-time use, including temporal framing, data preparation, and changing operational conditions.

Exploring the role of mass immunisation in influenza pandemic preparedness: A modelling study for the UK context.

The nature and timing of the next influenza pandemic is unknown. This makes it difficult for policy makers to assess whether spending money now to prepare for mass immunisation in the event of a pandemic is worthwhile. We used simple epidemiological modelling and health economic analysis to identify the range of pandemic and policy scenarios under which plans to immunise the general UK population would have net benefit if a stockpiled vaccine or, alternatively, a responsively purchased vaccine were used. Each scenario we studied comprised a combination of pandemic, vaccine and immunisation programme characteristics in presence or absence of access to effective antivirals, with the chance of there being a pandemic each year fixed. Monetarised health benefits and cost savings from any influenza cases averted were set against the option, purchase, storage, distribution, administration, and disposal costs relevant for each scenario to give a discounted net present value over 10 years for planning to immunise, accounting for the possibility that there may be no pandemic over the period considered. To support understanding and exploration of model output, an interactive visualisation tool was devised and made available online. We evaluated over 29 million combinations of pandemic and policy characteristics. Preparedness plans incorporating mass immunisation show positive net present value for a wide range of scenarios, predominantly in the absence of effective antivirals. Plans based on the responsive purchase of vaccine have wider benefit than plans reliant on the purchase and maintenance of a stockpile if immunisation can start without extensive delays. This finding is not dependent on responsively purchased vaccine being more effective than stockpiled vaccine, but rather is driven by avoiding the costs of storing and replenishing a stockpile.