Results

Table 1. Charasteristics of patients at admission in rehab centers
vars Overall
n 96
age_event (mean (SD)) 43.6 (20.8)
sex = M (%) 69 (71.9)
days_icu (median [IQR]) 27.0 [20.0, 35.0]
days_rehab (median [IQR]) 72.0 [43.8, 128.2]
tracheo_adm = 1 (%) 59 (61.5)
breath_adm (%)
Autonomous 75 (78.9)
Autonomous + O2 18 (18.9)
Mechanical 2 (2.1)
feed_adm (%)
Oral 36 (37.9)
NG-tube 41 (43.2)
PEG 18 (18.9)
urinary_cateth_adm = 1 (%) 91 (94.8)
bedsore_adm = 1 (%) 31 (32.3)
craniolac_adm = 1 (%) 23 (24.0)
Table 2. Diagnosis, CRS, DRS and nr. of CIRS comorbidity at admission and dismission from rehab
vars Admission Dismission p
n 96 96
diagnosis (%) <0.001
Death 0 (0.0) 11 (11.5)
Emersion 55 (57.3) 73 (76.0)
MCS 21 (21.9) 8 (8.3)
VS 20 (20.8) 4 (4.2)
crs (median [IQR]) 23.0 [7.0, 23.0] 23.0 [21.0, 23.0] 0.014
drs (median [IQR]) 18.0 [16.0, 21.0] 9.0 [5.0, 15.2] <0.001
totComorb (median [IQR]) 5.0 [3.8, 6.0] 2.0 [1.0, 3.0] <0.001
Figure 1. Mechanisms of Traumatic Brain Injury (TBI)

Figure 1. Mechanisms of Traumatic Brain Injury (TBI)

Figure 2. Lesions revealed at imaging

Figure 2. Lesions revealed at imaging

Figure 3. CIRS measured comorbidity rate at admission and after rehabilitation

Figure 3. CIRS measured comorbidity rate at admission and after rehabilitation

Figure 4. Admission - Dismission DRS delta vs Age at TBI (A) and nr. of severe comorbidities (CIRS >= 2; B)

Figure 4. Admission - Dismission DRS delta vs Age at TBI (A) and nr. of severe comorbidities (CIRS >= 2; B)

Figure 4. Admission - Dismission DRS delta vs Age at TBI (A) and nr. of severe comorbidities (CIRS >= 2; B)

Figure 4. Admission - Dismission DRS delta vs Age at TBI (A) and nr. of severe comorbidities (CIRS >= 2; B)

Table 3. Logistic model for association with risk of failed functional improvement
vars OR CI95 p
age_event 1.026 0.994 - 1.059 0.118
sexM 0.955 0.265 - 3.442 0.944
diagnosis_admMCS 2.460 0.558 - 10.855 0.235
diagnosis_admVS 2.141 0.555 - 8.261 0.269
tot_comorb_rehab 1.663 1.089 - 2.538 0.018
Figure 5. ROC curves for Machine Learning models tested: Random Forest (grey), Lasso regression (black solid) and SVM with Polynomial kernel (black dotted)

Figure 5. ROC curves for Machine Learning models tested: Random Forest (grey), Lasso regression (black solid) and SVM with Polynomial kernel (black dotted)

Table 4. Machine Learning models predictive performance
models auc CI sensitivity specificity ppv npv
RF 0.860 0.819 - 0.9 0.835 0.711 0.917 0.529
Lasso 0.826 0.77 - 0.882 0.751 0.856 0.952 0.472
SVM Poly 0.775 0.715 - 0.835 0.759 0.722 0.913 0.439