ICB-led Applied Research Projects
These research projects are led by ICB Colleagues supported by researchers to target ICB priority areas.
Finding projects: Use “Ctrl + F” to use the find function on your browser. Then use key terms to seek projects in the topic of your interest. It is best to try several alternative words e.g. “Birth” or “Maternity” or “Maternal” or “natal”.
Improving uptake of simulation in healthcare: User-driven development of an open-source tool for modelling patient flow
The PathSimR model is a versatile simulation model purpose-built in BNSSG for modelling patient pathways in healthcare. Providing a free and flexible solution, the software has been used in BNSSG and further afield in other NHS systems for various projects and pieces of work. This paper tells the story of how the software was developed and provides full details on its workings and functionality.
Balancing capacity and the flow of patients between acute hospital and community care is a difficult planning problem, involving the consideration of uncertain patient arrivals and variable lengths of stay. Through various scenarios, our modelling has helped reveal the optimal allocations of capacity along this pathway. The modelling was undertaken in the immediate months following COVID-19.
Out-of-area placements occur when there is no capacity available to satisfy the demand within the local area. This can often happen for high acuity mental health care, with patients sent potentially large distances to other facilities. Our modelling, using the PathSimR model, considered various capacity-side scenarios in order to mitigate such eventualities.
Elective wait lists had increased significantly following the COVID-19 pandemic and, in the early stages of recovery from the pandemic, it was unsure how much of the many ‘missed referrals’ would return. Modelling was undertaken locally to understand the possible size of the waiting list and waiting times should varying proportions return. The model was also applied at national England-wide level.
Our PathSimR model was used to model the future-state centralised stroke pathway planned in BNSSG. The modelling involved the calibration of the pathway model and its use to answer questions around how much flexible capacity would be required at various times, in order to ensure that the vast majority of patients would encounter no delay on admission to the hyper-acute stroke unit.
This paper reports on the modelling approach taken in BNSSG to estimate future wait list size at trust and specialty level, based on different assumptions regarding future demand and capacity levels. Being simple and scalable, the model has since been applied to every hospital trust and specialty in England, with such projections updated on a monthly basis.
The False Economy of Seeking to Eliminate Delayed Transfers of Care: Some Lessons from Queueing Theory
This work challenged the established wisdom that it is necessary in well-performing healthcare systems to “eliminate” delayed transfers of care (sometimes referred to as ‘bed blocking’). The study, using methods from the mathematical discipline of queueing theory, found that pursual of such a policy is likely to be uneconomical, as it would require large amounts of community capacity to accommodate even the rarest of demand peaks, leaving much capacity unused for much of the time.
Implementation of the Recommended Summary Plan for Emergency Care and Treatment (ReSPECT)
The ReSPECT process is an initiative that creates personalised recommendations for a person’s clinical care and treatment in a future emergency in which they are unable to make or express choices. This initiative was implemented in the local area during the pandemic. The aim of the analysis is to determine the equity of the ReSPECT form implementation process (during the first Covid-19 wave) and any associated changes to how patients and their local health care bodies interact thereafter. This will help inform future commissioning decisions regarding the use of the ReSPECT form.
Using hyper-local population health management to increase the number of people getting vaccinated against COVID-19
IBNSSG ICB commissioned local campaigns to encourage people who are less likely to get vaccinated, to get the COVID-19 vaccine. This analyses will help us understand how effective these campaigns were so we can improve how we organise local and national health programmes involving large numbers of people in the future.
P-NEWS: personalised early warning scores for critical admission patients
This project aims to reduce intensive care admissions by intervening earlier to correct problems before they become critical. They plan to do this by using patient observations and advanced analytics to produce an accurate risk of deterioration for any individual patient. The National early Warning Score (NEWS) is a “one size fits all” score highlighting how sick a patient is. Unfortunately, NEWS does not take into account important features such as the diagnosis and past medical history. This project aims to refine the score and predict deterioration sooner.
This project is part of the Health Data Research UK South Better Care Partnership.
Segmenting the population into particular groups based on individual attributes and/or healthcare activity is a key component of Population Health Management (PHM). However, there exists a multitude of possible methods to perform population segmentation, each with their own pros and cons. This project reviewed 16 of the most commonly used approaches in determining which method is most appropriate for answering particular types of question. Findings have since informed our choice of segmentation method when working on projects within our PHM programme.
Referral to treatment (RTT), measuring the proportion of waiting patients waiting under 18 weeks, is the principal barometer of elective performance in the NHS, and is used to monitor just how long patients are waiting for planned treatment. For healthcare systems, it is important to understand and model the dynamics of the RTT pathway, so that future waiting times can be reliably projected and the effect of changes in referrals and capacity can be assessed. Our computer simulation model has been used regularly to such end, both for different hospital trusts and clinical specialties.
Without sufficient capacity, clinical pathways can become blocked, with patients ready for discharge but unable to be transferred downstream. This is both bad for patients as well as hospitals. However, estimating the optimal required capacity is not straightforward. While spreadsheet approaches are quick and easy, they typically under-estimate the required number of beds to commission. Here, we develop a more robust approach in providing a customisable and reusable computer model, which is applied to estimating capacity for the future stroke pathway.
Population Health Management to identify and characterise ongoing health need for high-risk individuals shielded from COVID-19: a cross-sectional cohort study
In the early stages of the COVID-19 pandemic, approximately 30,000 vulnerable BNSSG residents were asked to ‘shield’ in order to protect themselves from the dangers of COVID-19 infection. However, little was known about this cohort of individuals. Using linked data, six distinct segments were identified within the shielding population. Awareness of these helped us to better tailor advice to patients and support local primary care teams in managing their conditions while shielding.
Right at the start of the pandemic, there was very little information to help managers and clinicians understand the amount of intensive care beds needed to match the possible incoming demand. This was important as it was difficult to convert beds to intensive care specification. Yet, if too few were converted then this could result in patients not able to access the level of care they required. To address this issue, a computer simulation model of COVID-19 patient flow was rapidly developed and used as part of the crucial initial response to the high numbers of cases in Spring 2020.
While the immediate effects of COVID-19 were on emergency hospital care, it quickly became apparent that the decision to postpone elective treatment in Spring 2020 would have a severe impact on waiting times. The questions were, how much of an effect would this have, and how quick could we recover? To answer these, an existing tool used for modelling referral to treatment (RTT) dynamics was recalibrated and used to project waiting times under various scenarios considered plausible at the start of the pandemic.
With bed occupancy rapidly increasing in the second wave of the pandemic, hospital planners required estimates of the likely number of admissions in the coming days. A simple time series forecasting model was built and deployed for daily use in projecting acute and intensive care bed occupancy for all local hospitals. This helped to ensure that the appropriate number of beds were prepared and new infection wards were opened as required.
The impact of increased outpatient telehealth during COVID‐19: Retrospective analysis of patient survey and routine activity data from a major healthcare system in England
In order to help limit hospital infections, a significant amount of outpatient consultations were moved from the physical to the virtual setting in the early stages of the pandemic. In examining a large number of patient surveys, it was found that more respondents ‘preferred’ virtual than physical appointments with seven times as many finding them ‘less stressful’ than ‘more stressful’. Results have helped to inform the possible suitability of video consultations going forward.
While the NHS has thankfully not had to introduce triage for intensive care admissions during the first year of the pandemic, it had got close at times. When demand for such a resource outstrips supply, it is arguably important to promote better access to those who have most to benefit. However, there is little evidence to support just how much can be gained by implementing triage. Our work addressed this gap, in finding that triage can reduce total life-years lost by 12%, should demand ever exceed supply.
Vaccination centres were critical for scaling up mass vaccination of the population against COVID-19. Yet planners had very little information to guide the configuration of these sites, which had to be set up in a matter of weeks. At the Bristol Ashton Gate site, computer simulation modelling was used to inform the maximum throughput of the centre, in terms of the number of people who could be vaccinated each day. Model outputs were used for the crucial opening months of operation.
Projecting the effect of easing societal restrictions on non‐COVID‐19 emergency demand in the UK: Statistical inference using public mobility data
While much of the focus had been on COVID-19 cases, societal restrictions also had a significant impact on non-COVID-19 hospital admissions. With less sporting injuries and road accidents, for instance, there was less emergency demand at local hospitals. In the early months of 2021, with lockdown gradually being relaxed, a regression model was used to forecast, based on expected increases in public mobility, the extent to which bed admissions could rise.
A longstanding, and broadly unchallenged, result states that hospitals should target 85% average bed occupancy in balancing the risks to patient safety of too little capacity with the financial consequences of too much. However, a single measure is insensitive to the range of conditions that realistically exist on the ground. In producing a ‘look up’ table based upon ward size and specialty, our modelling reveals a set of more accurate targets that can be used by hospital managers and commissioners.
During the first wave of the pandemic, there was much uncertainty regarding the amount of ‘pent up’ demand for mental health services that could lead to pressure following release of lockdown. A versatile discrete-time queuing model was quickly produced and used to examine the potential effect of a number of different demand trajectories and service interventions designed to mitigate severe system pressure.
Establishing an SEIR-based framework for local modelling of COVID-19 infections, hospitalisations and deaths
While epidemiological modelling has been routinely used to inform decision making at the national level, there has been very little to guide planning at the local level. Ultimately, the key measure of interest has been the expected future number of acute COVID-19 admissions. In using a compartmental ‘SEIR’ type model, a cross-system multi-disciplinary working group was set up to configure various plausible scenarios, with outputs shaping the local response in determining how many beds were needed to safely accommodate the future demand.
Building on previous work involving a compartmental ‘SEIR’ type model, a number of technical enhancements were made in order to take account of the effect of vaccination on transmission dynamics. The model was then used to examine a number of scenarios relating to the lockdown relaxation roadmap of early 2021. Projections were validated with the Autumn 2021 rebound in hospital cases falling comfortably within the modelled interquartile range.
As well as the immediate impact of COVID-19 infection on acute healthcare services, there had been a concern that the longer-term effects (so-called ‘Long-COVID’) could be placing additional demands on other healthcare settings. The BNSSG System Wide Dataset was used to identify evidence of statistically significant increases in healthcare activity within three months of a COVID-19 diagnosis.