Table 2.
Categorical Groupings of NIHSS to Best Discriminate 30-Day Mortality Risk
| No. of Groupings | Risk-Tree Cut Points for the Groupings | Chi-Square for 30-Day Mortality | c-Statistic (95% Confidence Intervals) |
|---|---|---|---|
| 2 | 0–13, 14–42 | 5378 | 0.74 (0.74–0.75) |
| 3 | 0–13, 14–21, 22–42 | 6069 | 0.75 (0.75–0.76) |
| 4 | 0–7, 8–13, 14–21, 22–42 | 6418 | 0.80 (0.79–0.80) |
| 5 | 0–7, 8–13, 14–18, 19–21, 22–42 | 6541 | 0.80 (0.79–0.81) |
| 6 | 0–7, 8–13, 14–18, 19–21, 22–27, 28–42 | 6651 | 0.80 (0.79–0.81) |
| Original (NIHSS as continuous variable) | 0, 1, 2, 3, 4, …., 42 | 6856 | 0.82 (0.81–0.83) |
NIHSS indicates National Institutes of Health Stroke Scale.
Classification and regression tree methodology (in R) with predetermined No. of nodes were applied to identify the categories of NIHSS. Larger chi-square statistic and c-statistic closer to 1.0 indicates more optimal discrimination of risk.