Table 3.
B coefficient | OR (95% CI) | P-value | Sensitivity, % | Specificity, % | Positive predictive value, % | Negative predictive value, % | Yi | AUC | |
---|---|---|---|---|---|---|---|---|---|
Univariable model | 78.6(63.2–89.7) | 63.3(56.8–69.4) | 27.5(23.2–32.3) | 94.3(90.3–96.8) | 41.9 | 0.77(0.70–0.85) | |||
ADL (BI, range 10–100p), per 5 gained points | −0.04 | 0.95(0.94–0.97) | <0.001 | ||||||
Multivariable model | 75.0(58.8–87.1) | 71.9(65.6–77.6) | 31.6(26.0–37.6) | 94.3(90.6–96.6) | 46.9 | 0.80(0.72–0.87) | |||
ADL (BI, range 10–100p), per 5 gained points | −0.04 | 0.96(0.94–0.98) | <0.001 | ||||||
Cognitive function (MoCA, range 3–30p), per 1 gained point | −0.06 | 0.94(0.87–1.01) | 0.10 | ||||||
Stroke severity (NIHSS, range 0–19 p), per 1 gained point | 0.03 | 1.03(0.92–1.15) | 0.62 | ||||||
Age (range 19–97), per 1 gained year | 0.04 | 1.04(1.01–1.08) | 0.02 | ||||||
10 – fold cross-validation (90% training set and 10% testing set), mean ± S.D. | |||||||||
Sensitivity | Specificity | Positive predictive value | Negative predictive value | Yi | AUC | ||||
Univariable model | 0.57 ± 0.15 | 0.70 ± 0.40 | 0.22 ± 0.17 | 0.95 ± 0.06 | 0.27 | 0.76 ± 0.11 | |||
Multivariable Model | 0.70 ± 0.09 | 0.69 ± 0.32 | 0.29 ± 0.19 | 0.94 ± 0.06 | 0.39 | 0.80 ± 0.12 |
The results of tuned univariable and multivariable logistic regression models and 10-fold cross validation performed per respective model. Yi, Youden's index (sensitivity (%) + specificity (%) – 100[or 1 for cross validated results]); ADL, personal activities of daily living; BI, Barthel index, the score in BI is increased with 5 points; MoCA, the Montreal Cognitive Assessment; NIHSS, the national institutes of health stroke scale; BI and MoCA were administered 36–48 h after admission to the stroke unit. NIHSS, hospital admission point; BI and MoCA, a higher score indicates better outcome; NIHSS, a higher score indicates worse outcome.