A total of 1635 patients were admitted to the NSICU during the study period, among which 329 patients had ICU admission more than 24 hours. After 148 patients were excluded[incomplete information (n = 21), aged < 18 years old (n = 68), palliative care (n = 36), discharged against advice (n = 16), pupil disease(n = 7)], the rest 181 patients were included in the analysis(Fig. 1).
As depicted in Table 1., the patients included in this study had an mean GCS score of 5 (3–9), an mean GCS-Pupils score of 5 (1–10) and an mean FOUR score of 7 (0–16); the average age was 60 (31–89); 98 were male (54.1%); an mean APACHE II score of 19.4 ± 4.6. Regarding the primary disease causing neurological function impairment, there were 66 cases (36.5%) of cerebral haemorrhage, 46 cases (25.4%) of ischaemic-hypoxic encephalopathy, 26 cases (14.4%) of craniocerebral trauma, 24 cases (13.3%) of acute cerebral infarction, 15 cases (8.3%) of subarachnoid haemorrhage, and 4 cases (2.1%) of central nervous system infection. Co-morbidities were hypertension in 126 cases (69.6%), diabetes mellitus in 73 cases (39.8%), and atrial fibrillation in 23 cases (12.7%). Only 5 cases without chronic disease (2.8%).
Of these patients, 180 (99.4%) were under tracheal intubation/ tracheostomy status. Patients in this study had long period of MV time (10 (1–31),day), ICU LOS (13 (8–30), day) and hospital LOS(41 (8-100),day)), a poor prognosis (6-month GOSE(2 (1–3) ) and poor survival indicators (28-day mortality 29.3%, in-hospital mortality 36.5% and 6-month mortality 40.3%).
Cox logistic regression was adopted to conduct the risk factors of 6-month GOSE, the results suggested that the GCS scores(p = 0.00, 95%CI 0.176 ~ 0.395), the GCS-Pupils scores(p = 0.00, 95%CI 0.264 ~ 0.476) and the FOUR socors (p = 0.00, 95%CI 0.427 ~ 0.608) were the risk factors. Then comparative analysis was conducted on the predictive value of the three scales.
Correlation analysis results between the scores of the three scales and the primary and secondary outcomes
As depicted in Fig. 2, the scores of all the three scales were correlated with the primary endpoint(6-month GOSE (P < 0.01)), and FOUR had the best predictive value(OR = 0.5, 95%CI: 0.4 ~ 0.5). The receiver operating characteristic curve (ROC) was graphed then(Fig. 3.), the results revealed that all the three scores had good predictive accuracy. No significant difference was found in AUROC and 95% CI, the FOUR was superior to the other two scales in terms of cutoff value with a cutoff value of 11 points, while the GCS and GCS-Pupils were 8 points.
In the comparison of the secondary endpoints, the scores of the three scales were associated with the mortality outcomes(P < 0.01), and the FOUR scores had the best predictive value (28-day mortality (OR: 0.5, 95%CI: 0.4 ~ 0.5), in-hospital mortality (OR: 0.6, 95% CI: 0.5 ~ 0.7), 6-month mortality (OR: 0.5, 95% CI: 0.4 ~ 0.6). Only the GCS-Pupils scores(P = 0.01) and the FOUR scores(P < 0.01) were correlated with hospital LOS, and FOUR had the best predictive value(OR:1.3, 95% CI:1.0 ~ 2.4). Only the FOUR scores had predictive value for MV time(P < 0.01, OR: 0.9, 95% CI: 0.1 ~ 1.2). None of the three scales had predictive value for ICU LOS (P > 0.05).
The results demonstrated that the GCS-Pupils and the FOUR were superior to the GCS. Subgroup analysis was conducted to compare the predictive value of the two scales on neurological function impairment for more severe patients(grouped based on GCS score ≦ 8).
Comparison of the predictive value of the GCS-Pupils and the FOUR on neurological function impairment in the subgroup analysis of more severe patients
In this study, a total of 147 patients had a GCS score ≦ 8. As depicted in Fig. 4, the scores of the FOUR were correlated with the primary outcome(6-month GOSE: OR 2.3, 95%CI 1.6 ~ 3.3, p < 0.01), the mortality outcomes(28-day mortality: OR 9.1, 95%CI 4.5 ~ 26, p < 0.01, the in-hospital mortality: OR 6.9, 95%CI 3.9 ~ 15.5, p < 0.01 and the 6-month mortality:OR 1.6, 95%CI 1.4 ~ 1.9, p < 0.01) and the hospital LOS(OR 6.8, 95%CI 0.4 ~ 13.6, p = 0.04), but no correlation with ICU LOS and MV time(p > 0.05).
In the severe subgroup analysis, the GCS-Pupils had no correlation with the primary and secondary outcomes compared to the FOUR(p > 0.05).