Table 3 Linear regression results highlighting univariate associations between sample size (categorized as <100, 100–500, 500–1000, >1000), journal impact factor (categorized as <3, 3–5, 5–10 and >10) and publication year with APPRAISE-AI overall scores

From: Predicting outcomes after moderate and severe traumatic brain injury using artificial intelligence: a systematic review

Variable

Mean difference in APPRAISE-AI overall score

95% confidence interval

p value

R2

Univariate linear regression

Impact factor [Reference IF < 3]

0.026

0.21

 IF 3–4.9

5.8

−1.5–13.1

 IF 5–9.9

8.1

−0.97–17.1

 IF > 10

15.5

4.5–26.5

Sample size [Reference: <100 patients]

0.002

0.31

 100–499 patients

8.4

−1.0–17.7

 500–1000 patients

10.3

0.2–20.4

 >1000 patients

18.3

8.4–28.2

Publication year

0.65

0.04–1.27

0.034

0.11

Multivariable linear regression

Impact factor [Reference IF < 3]

0.041

0.65

 IF 3–4.9

0.49

−5.9–6.9

 IF 5–9.9

5.9

−2.6–12.3

 IF > 10

10.4

0.8–20.0

Sample size [Reference: <100 patients]

<0.001

 100–499 patients

11.4

3.6–19.1

 500–1000 patients

8.8

0.09–17.6

 >1000 patients

18.7

10.9–26.6

Publication year (per 1-year increase)

0.70

0.23–1.16

0.002

Country of data collection [high-income country with reference to upper middle-income country]

8.6

1.7–15.4

0.011

  1. Multivariable linear regression model was additionally adjusted for country of data collection (high-income compared to upper-middle income). p values were determined for variables using likelihood ratio testing (to determine overall association, rather than association per variable level). Publication year ranged from 1997 to 2024 in the study. In the regression model, publication year was kept as a continuous variable and centered around 0. Bolded values represent statistically significant associations with p < 0.05.
Back to article page