The updated set of SSNAP Key Indicators (KIs) introduced for October 2024 include 4 thrombolysis measures in the Reperfusion domain (domain 3). These are:
- Percentage of all stroke patients given thrombolysis
- Percentage of stroke patients with extended indicators for thrombolysis given thrombolysis
- Percentage of stroke patients given thrombolysis (within 4h) compared with bespoke site-specific target
- Median time between clock start and thrombolysis
KI 3.3 is based on machine learning using the SSNAP dataset, considering over 250,000 stroke admissions over three years of which approximately 26,000 patients were thrombolysed. The new KI and its supporting information provide sites with their own bespoke estimate of their ‘best achievable’ thrombolysis rate (bespoke site-specific target) for patients presenting to hospital with a known or best estimate of an onset time within the last four hours. Allowing for patients treated up to 4.5 hours after onset. It does not include patients with ‘extended indications’ for thrombolysis beyond 4.5 hours or where the onset time is unknown.
How is the ‘best achievable’ thrombolysis rate calculated?
The developmental research work for this indicator came from SAMueL – Stroke Audit Machine Learning and the work is published in this paper.
The most important feature of this bespoke site-specific target is that it is based on a site’s own local population and calculated from each site’s own thrombolysis decision making over the preceding three years. This avoids the issue that frequently arises with comparative audit, in which sites may attribute differences in their own local populate for any observed differences in patient selection or eligibility for thrombolysis. SAMueL showed that less than half of the observed variation between sites is accounted for by differences in demographics or other patient level characteristics and the three main contributors to between-site variation in thrombolysis use are:
- Differences in the reporting of a ‘known or best estimate’ for the onset time;
- Differences in door-to-needle time;
- Differences in willingness to thrombolyse patients with ‘less than ideal’ clinical characteristics, particularly lower stroke severity, higher pre-stroke disability, older age and presentation late in the 4.5-hour treatment window.
Based on a site’s own local population, machine learning allows SSNAP to calculate the thrombolysis rate that could be achieved at each site if:
- The site achieved an upper-quartile proportion of patients with a known or best estimate for the onset time;
- The site consistently achieved a door-to-needle time of 30 minutes;
- The site replicated the willingness to thrombolyse ‘less than ideal’ patients that would be observed if those same patients presented to a site with a ‘top quartile’ thrombolysis rate.
The impacts from each of these three pathway improvements are additive, and when they are combined they produce a best achievable thrombolysis rate which can be directly linked to the number of additional patients predicted to recover with no or minimal disability (a modified Rankin scale score of 0-1) at 90 days. The SAMueL analysis included health economic measures that showed that sites who were closer to their best achievable thrombolysis rate were still delivering cost savings from reduced lifetime disability, addressing the objection that additional patients treated at higher-thrombolysing sites are those with lesser degrees of stroke and a good natural prognosis even without treatment.
Interpreting SSNAP reports
SSNAP reports will contain the following measures that contribute to key indicator 3.3:
How can sites use their SSNAP reports to support quality improvement in thrombolysis?
The figures reported in SSNAP are derived from machine learning on the acute dataset for the last three years. Targets are available here: The target for each site will be updated annually following publication of that year’s annual results.
Local and regional quality improvement using these data is supported by the SAMueL WebApp, which is available at https://stroke-predictions.streamlit.app/. On the App, clinical and audit staff can:
- On the ‘Descriptive Statistics’ tab, use their own unique site code (known only to their site) to investigate the sources of difference in their own local population from the national averages or from other sites elsewhere in the country (data used in the App is restricted to England and Wales);
- On the ‘Pathway Improvement’ tab, examine the impact on thrombolysis rates and disability outcomes from improvements to the thrombolysis pathway, and how those changes would affect their site’s relative position in the distribution of thrombolysis rates across England and Wales;
- On the ‘Thrombolysis Decisions’ tab, staff in clinical/audit or case review meetings can compare how patients presenting with various clinical characteristics and pathway timings would be treated if they presented to another hospital – either a ‘benchmark’ site in the top quartile for ‘willingness to thrombolyse’, or a large group of ‘average’ hospital sites.
These functions are not designed to provide decision-making guidance in the management of individual patients in the acute situation. They are intended to support analysis and discussion of decision-making in audit settings, and to provide insights into how patients would be treated if they were to present elsewhere – allowing comparison between sites based on what happened to real patients in the preceding three years. The App can help sites to identify where the greatest disability gains may be obtained from changes to the reperfusion pathway and avoid quality improvement effort being wasted on other changes that will have only limited impact on outcomes.