Predictors of driving outcome after traumatic brain injury

Role of premorbid factors in predicting safe return to driving after severe TBI

“…Similar results have been found for demographic variables that did not show any significant correlation with the open-road outcome [17,[19][20][21].…”

“…Korteling and Kaptein [20] explained the 35.3% of variability in rated driving performance combining two laboratory neuropsychological tests ('perceptual speed' and 'tracking reaction') with coma duration and reported driving experience. Finally, Coleman and colleagues [21] were able to correctly classify 80.3% of the subjects (r 2 ¼ 0.53) between drivers and non-drivers using neuropsychological tests, an index of patients' awareness of their own deficits, and two other subjective indexes on perception of the patients' ability to drive safely.…”

“…Unfortunately, the results of various studies in predicting driving outcome from different predriving parameters are highly inconsistent, and range from a reported predictive power of about 20% to 94% of explained variance (the predictive power refers to the proportion of variance of the outcome measure that can be explained by predictor variables) [5,[9][10][11][15][16][17][18][19][20][21][22][23][24][25]. For instance, Galski and colleagues evaluated the predictive power of 21 physical and neuropsychological tests thought to assess visuo-perceptive, motor, and cognitive skills relevant for safe driving, against a behind-the-wheel evaluation divided into 26 different tasks [16].…”

“…Other studies have purported to account for about 40% to 50% of the variance in the driving outcome measure [19][20][21]. In Sivak's et al research [19], which studied the role of perceptive and cognitive skills on driving, the 'picture completion test' yielded the strongest correlation (r ¼ 0.72; r 2 ¼ 0.52) with a 'composite driving index' derived from an on-road assessment of five different aspects of the patients' performance.…”

Editorial

“…Unfortunately there is no standard procedure or agreement, and predictive power is estimated among authors to vary widely from 40% to 93% [1][2][3].…”