Is the ISAN score valid against stroke datasets with high levels of completeness?

Published Date: 21st November 2016

Publication Authors: Hill AM

Parallet Session 2A: High Scoring Abstract

Introduction
In 2015, Smith et al analysed the Stroke Sentinel National Audit Programme dataset to produce a risk assessment tool for predicting 7 and 30 day mortality and the prevalence of stroke associated pneumonia (SAP). In our study we analysed our Trust’s stroke population over the last year using the ISAN score. 1 key feature of our dataset is that we have high rates of NIHSS completion in comparison to the national data used to produce ISAN. In the ISAN study cohort, the ‘excluded’ cohort with unknown NIHSS had much higher mortality than the included group and it was predicted that this underestimated mortality and SAP prevalence in a real-world population with high rates of data completion

Method
ISAN scores were calculated for 1 year’s data for all stroke patients recorded by our Trust on the Stroke Sentinel National Audit, and patients risk-stratified accordingly. 7-day, 30-day and SAP prevalence was calculated for each risk-stratified group. Population characteristics for our cohort were calculated and compared to the ISAN study cohort, and odds ratios for each risk factor calculated. Original study methodology and calculations were reproduced to enable direct comparison to the original populations

Results
Data completion rates in our population were high (97% had complete NIHSS scores). 7-day and 30-day mortality risks were statistically significantly higher. Stroke associated pneumonia prevalence was higher but did not reach statistical significance. Our cohort included a higher proportion of haemorrhage strokes (13% versus 8% in original study); other characteristics were similar. Predictors of pneumonia showed similar trends in both groups with some key differences. Gender did not appear to be a significant factor (Male ¼ OR 0.88 vs 1.23 in original study). Age was a less strong predictor in our population. Pre-admission diabetes was a much stronger predictor of pneumonia (OR 1.47 vs 0.97); whilst pre-stroke mRS > 2 was less strong (OR 1.60 versus 2.80).

Conclusion
ISAN demonstrated a good predictor of pneumonia risk in our population. However, the excluded population in the data used to calibrate the data had high mortality and may have excluded a significant population with stroke associated pneumonia. Local calculations with a highly complete dataset showed higher mortality, SAP prevalence, intracerebral haemorrhages, and lower risk from male gender or pre-stroke mRS > 2. Recalibration with a complete dataset is recommended with a highly-complete dataset; any national comparison of SAP utilising the SSNAP dataset would need to take into account data completeness rates to estimate the ‘gap’ in missing data

 

Hill, A. (2022). Is the ISAN score valid against stroke datasets with high levels of completeness?. International Journal of Stroke. 11 (Suppl 4), 9

 

« Back