This website uses cookies to improve your experience, deliver our services, and the anonymised analysis of our website usage. To opt out of analytical cookies select 'Allow only essential cookies'. Please read our cookie policy   
UAT
Case studies > 2022 annual report > Using SSNAP data for pathway optimisation/modelling

Using SSNAP data for pathway optimisation/modelling

Dorset County Hospital NHS Foundation Trust

Clinical challenge
The opportunity arose for pathway redesign in our local service, with an opportunity to refine our bed base (hyperacute, acute, rehabilitation), ambulatory care service and to develop our ESD service into a future Integrated Community Stroke Service (ICSS).

Solution
Data:
  1. 4 years of SSNAP data were pulled and analysed for seasonal trends, and average stroke numbers/yr (Exporting SSNAP Data
  2. Patients were stratified using mRS (0-6), and length of stay (LOS) was analysed for each category
  3. Stakeholders agreed the ideal LOS per mRS category in comparison to average existing LOS
  4. Proportions of patients who fell within the ideal LOS per group was reviewed to ensure ideals were reasonable
  5. Statistical Process Control  charts were created from SSNAP export including all patients. Notes were pulled for;
    1. patients whose LOS was the same or below the ideal LOS for that group, to review the care received for patients with an ideal LOS and determine features of good practice that could be replicated.
    2. a number of patients who sat above the upper control, to understand what happened to patients with a significantly longer LOS than expected, why this happened, and whether this was avoidable or needed to be factored into bed modelling.
Assumptive bed modelling:
  1. Stakeholders agreed proportion of LOS for each mRS group which would likely be an acute and rehabilitation (medically stable) stay
  2. SSNAP data and hospital ‘midnight bed occupancy’ data was bought together by the business intelligence team to show current bed utilisation
  3. A factor was identified that when used in the model, reduced the average LOS for each mRS group’s stay, in line with the ideal LOS identified. The model also accounted for the variation created by patients whose LOS was longer, with the assumption that this was due to social care delays. This modelling also explored future bed utilisation, allowing 85% bed occupancy if the ideal LOS were realised for example by optimising and tailoring services to each patient (mRS) group.
Pathway optimisation:
  1. The pathway was considered for each patient group with assumptions made about the proportion of patients who could be seen in various settings (some of which would need to be developed) - for example ambulatory care, ESD, dedicated downstream rehab beds, community stroke team
Impact
This graphic shows the pathway, the patient numbers expected, mRS proportions assumed in the modelling, historic LOS and ideal LOS in an optimised pathway. This has been helpful in engaging with services to articulate a vision for the future- a vision for improvement.
 
Bed modelling outputs were used to inform movement of dedicated rehabilitation beds off the acute site to create a new stroke rehab unit in a community hospital during COVID. To support this, outputs were triangulated using the GIRFT bed modelling tool using population size and 3 x snapshot audit data from all patients on the ward. This investigated whether patients could be safely managed in a non-acute setting. All approaches supported the suggested bed numbers from the bed modelling process described above based on modelling SSNAP data. 
 
Rehab beds moved off site in August 2020. LOS reductions were seen in line with ideals suggested whilst maintaining patient outcomes. SSNAP data is used to continually monitor ourselves against the assumptions in the above model as well as the usual process measures collected by SSNAP re quality. We continue to use this model in the development of our Integrated Community Stroke Services.

Reflection
The project brought together the clinical team and business intelligence team to create a data informed model that met clinical standards. We were reassured by the use of real-life data, including all the ups and downs that happen on stroke units, accounts for seasonal and natural variation that was also triangulated with other audit methods.

Find us

Sentinel Stroke National Audit Programme
Kings College London
Addison House
Guy's Campus
London
SE1 1UL

Support

0116 464 9901
ssnap@kcl.ac.uk