Applying advanced data science for health and social care operational decision-support to reduces delays in care.

点击翻译按钮获取中文摘要

Using data science, considerable value can be gained from health/social care data for supporting operational, strategic, and clinical decision-making to improve care quality in terms of efficiency, patient outcomes, staff satisfaction and reduced costs [1,2]. For example, discrete-event simulation is a method which develops a computer model of a process or system, enabling experimentation with the model, rather than disrupting the real system. It offers high value for decision-support in healthcare, for example by identifying key levers for improvement in treatment pathways [3,4]. Due to the challenges in implementing these tools in an applied setting using real-time data for impact, few examples exist in the literature. Those that do focus on technical, rather than implementation challenges [5]. However with advancing technology and increasing interest from the public sector in the use of data science for decision-support, there is expected to be an increase in the development of real-time operational decision-support tools using simulation/AI in healthcare research [6]. My PhD took a design approach to these applications, emphasising an early focus on the challenges of implementation using iterative cycles of development and evaluation. However safe, effective, impactful applications also require advanced programming skills for modelling complex problems using free, open-source programming languages for shared learning. Combining formal training with applicaton to NHS problems will add value for health and social care partners, and will significantly advance the programming skills I gained during my PhD.A key aspect of this proposal is a 20% research component. This will involve engaging with NHS organisations in Bristol (BNSSG CCG) and researchers at University of Bath School of Mangagement to develop opportunities for using data science in the integrated health and social care space, and leveraging this with a future grant proposal. Engaging with PenPEG Exeter and Bristol Patient Safety Initiative will ensure that planned research is aligned with patient needs and current QI strategic goals. Discrete-event simulation will be used for pathway modelling between hospitals and social care providers; Random Forest methods will enable prediction of hospital lengths-of-stay for model inputs. Prototype modelling will conclude the feasibility of these methods toward a substantive research project focusing on delayed hospital discharges for patients, which can be attributed to lack of capacity in community-based care [7], and which impacts adversely on patients' mental and physical health [8]. Access has been granted through BNSSG CCG to NHS community health/social care services and other providers (acute hospital discharge planning, Local Authority social care provision) to map the problem situation holistically (how to optimise the balance of capacity in acute and community services to reduce delayed hospital discharges), and to access anonymised data for preliminary modelling. A CCG honorary contract has been approved, providing an invaluable opportunity to develop connections with healthcare and academic organisations, and to deliver the basis of substantive research which will support my future career goals. These ultimately include innovative and impactful applications, for example in early diagnosis, prevention and treatment of disease. This fellowship opportunity would additionally support publications from my PhD research, shifting academic discussions on real-time decision-support in healthcare toward real-world impact in the academic community. Conference presentations will disseminate new research, facilitating future collaborations in an under-researched area (community care) which has the potential for significant impact. Lack of availability of social care services is a recognised contributor to the pressures faced by hospitals; optimising capacity allocation can improve cost-efficiency and patient outcomes.