Heterogeneity in measurement of NIHSS: a systematic review and meta-analysis

Background: The National Institutes of Health Stroke Scale (NIHSS) is a pivotal clinical tool used to assess patients with acute stroke. However, substantial heterogeneity in the application and interpretation of stroke scale items can occur. This systematic review aimed to elucidate heterogeneity in measuring the NIHSS. Material and methods: A literature search was performed on PubMed/OVID/Cochran's CENTRAL from inception to 2023. The references of the included papers were reviewed for further eligible articles. Clinical characteristic, NIHSS values, and sources of heterogeneity were recorded. Non-human and non-English language articles were excluded. The study quality was assessed using MINORS and GRADE. Meta-analysis and meta-regression were performed using a random-effects model to explore the sources of heterogeneity. Results: Twenty-one papers for a total of 818 patients (mean per study: 39 ± 37) and 9696 NIHSS examinations (median per study: 8 [CI95% 2 to 42]) were included. Motor function had a higher ICC agreement (ranging from 0.85 [“Right Leg”] to 0.90 [“Right Arm”]) compared to the remaining items (ranging from 0.58 [“Facial Palsy”] to 0.85 [“Level of consciousness commands”]. The meta-regression showed a low effect size of covariates such as language version, remote evaluation, and retrospective analysis on NIHSS items (e.g., for “Level of consciousness commands,” language effect was 0.30 [CI95% 0.20 to 0.48] and for “Visual”, the retrospective assessment effect was -0.27 [CI95% -0.51 to -0.03]). Conclusion: The NIHSS scores showed moderate to excellent inter-rater agreement, with the highest heterogeneity in non-motor function evaluation. Using a non-English version, remote evaluation and retrospective analysis had small effects in terms of heterogeneity in the NIHSS scores.