12. ARTICLE (IN ENGLISH)
12.3 DISCUSSION
12.3.5 Future implications
12.3.5 Future implications
Although both prediction models perform well at predicting short term mortality after severe TBI, they are not perfect. The following candidate predictors should be tested for further improvement:
multiple trauma [59], pre-‐existing co-‐morbidities assessed with the Charlson score [60] and post-‐
injury complications such as pneumonia [61] and/or transfusion of platelets [62].
13. SYNTHÈSE
Cette étude de cohorte multicentrique a mis en évidence une relation robuste entre une description anatomique ou structurelle du TCC utilisant le HAIS et une description physiologique ou fonctionnelle utilisant le msGCS. Cette association a été observée chez les patients présentant une réactivité
la tomodensitométrie peut compliquer l'évaluation du HAIS. Troisièmement, alors que l’étude utilisait un modèle de prédiction de référence établi avec des patients plus jeunes (modèle IMPACT) [51, 52], l’âge médian plus élevé de notre cohorte introduisait un risque potentiel de biais de mixité.
Cependant, le modèle IMPACT avait déjà été validé dans un pays à revenu élevé avec une population âgée similaire [58]. Quatrièmement, le moment des évaluations des prédicteurs était légèrement différent entre les deux modèles de prédiction: le msGCS a été évalué à l'admission, tandis que le HAIS a été évalué jusqu’à 24 heures après l'admission. L’impact de cette différence de timing d’évaluation du prédicteur sur le résultat est difficile à estimer, mais a probablement une pertinence mineure. Cinquièmement, notre modèle de prédiction basé sur HAIS n'a pas été validé dans une cohorte externe. Nous pensons que le risque de surajustement est faible compte tenu des valeurs corrigées d'optimisme comparables d'AUROC pour les deux modèles de prédiction.
En conclusion, bien que ces deux modèles de prédiction permettent de prédire de manière adéquate la mortalité à court terme après un TCC sévère, ils ne sont pas parfaits. Pour être plus performants, les potentiels variables prédictives suivantes devront être évalués: les traumatismes multiples [59], les comorbidités préexistantes évaluées par le score de Charlson [60] et les complications post-‐
traumatiques telles que la pneumonie [61] et / ou la transfusion de plaquettes [62].
14. TABLE AND FIGURES
14.1 Figure 1 14.2 Figure 2 14.3 Figure 3a 14.4 Figure 3b 14.5 Figure 4a 14.6 Figure 4b 14.7 Table 1 14.8 Table 2 14.9 Table 3
Figure 1. Flow chart of enrolled and included patients.
Patients with inclusion criteria and consent n=921
10 deaths on arrival of OHEMS
Patients with predictive factors n=808
Patients with inclusion criteria and consent n=911
Patients with inclusion criteria and consent n=910
102 patients with missing predictive factors 1 death on scene after arrival of OHEMS
Figure 2a. Distribution of the categories of HAIS and motor GCS at ED.
0100200300400500Number of patients
020406080100Percent
4 5 6
HAIS
Subscale motor score 1-2 of GCS Subscale motor score 3-4 of GCS Subscale motor score 5-6 of GCS Number of patients
Figure 2b. Distribution of the categories of subscale motor score of GCS on ED and HAIS stratified in patients with normal pupil reactivity and abnormal pupil reactivity.
0100200300
Number of patients
020406080100Percent
4 5 6
HAIS
Subscale motor score 1-2 of GCS Subscale motor score 3-4 of GCS Subscale motor score 5-6 of GCS Number of patients
Normal pupil reaction
0100200300
Number of patients
020406080100Percent
4 5 6
HAIS
Abnormal pupil reaction
Figure 3. Accuracy of discrimination (AUROC) for the HAIS-based predictive model and the reference predictive model.
0. 00 0. 25 0. 50 0. 75 1. 00
Se nsi tivi ty
0.00 0.25 0.50 0.75 1.00
1-Specificity
HAIS-based predictive
model AUROC: 0.839 Reference predictive model AUROC: 0.826 Reference
Figure 4a. Calibration of the HAIS-based predictive model.
2.1
9.2
16.4
29.6
42.9
55.1 59.5
78.6
65.5
100.0
0 10 20 30 40 50 60 70 80 90 100
[0%-‐5%[ [5%-‐15%[ [15%-‐25%[ [25%-‐35%[ [35%-‐45%[ [45%-‐55%[ [55%-‐65%[ [65%-‐75%[ [75%-‐85%[ [85%-‐100%]
Observed death at 14 days (%)
Predicted death at 14 days (%)
Figure 4b. Calibration of the reference predictive model.
6.0 8.2
21.5 25.0
39.5
62.1
68.4 70.0 70.0
100.0
0 10 20 30 40 50 60 70 80 90 100
Observed death at 14 days (%)
Predicted death at 14 days (%)
15. REFERENCES
traumatic brain injury-‐related deaths-‐-‐United States, 1997-‐2007. MMWR Surveill Summ. 2011;60:1-‐32.
[5] Langlois JA, Rutland-‐Brown W, Wald MM. The epidemiology and impact of traumatic brain injury:
a brief overview. J Head Trauma Rehabil. 2006;21:375-‐8.
[6] Majdan M, Plancikova D, Brazinova A, Rusnak M, Nieboer D, Feigin V, et al. Epidemiology of traumatic brain injuries in Europe: a cross-‐sectional analysis. Lancet Public Health. 2016;1:e76-‐e83.
[7] Walder B, Haller G, Rebetez MM, Delhumeau C, Bottequin E, Schoettker P, et al. Severe traumatic brain injury in a high-‐income country: an epidemiological study. J Neurotrauma. 2013;30:1934-‐42.
[8] Langlois JA, Marr A, Mitchko J, Johnson RL. Tracking the silent epidemic and educating the public: enrollment: an IMPACT analysis. J Neurotrauma. 2007;24:270-‐80.
[17] Balestreri M, Czosnyka M, Chatfield DA, Steiner LA, Schmidt EA, Smielewski P, et al. Predictive
[23] Murray GD, Butcher I, McHugh GS, Lu J, Mushkudiani NA, Maas AI, et al. Multivariable Receiver Operating Curve Indices and Bayesian Network Analysis. PLoS One. 2016;11:e0158762.
[28] Gennarelli TA, Wodzin E. AIS 2005: a contemporary injury scale. Injury. 2006;37:1083-‐91.
[29] JD S. Rating the severity of tissue damage. I. The abbreviated scale. JAMA. 1971;215:277-‐80.
[30] Ringdal KG, Skaga NO, Hestnes M, Steen PA, Roislien J, Rehn M, et al. Abbreviated Injury Scale:
not a reliable basis for summation of injury severity in trauma facilities? Injury. 2013;44:691-‐9.
[31] Tohme S, Delhumeau C, Zuercher M, Haller G, Walder B. Prehospital risk factors of mortality and impaired consciousness after severe traumatic brain injury: an epidemiological study. Scand J Trauma Resusc Emerg Med. 2014;22:1. predicting outcome after traumatic brain injury. J Trauma. 2007;62:946-‐50.
[35] Demetriades D, Kuncir E, Brown CV, Martin M, Salim A, Rhee P, et al. Early prediction of brain injury: from prophecies to predictions. Lancet Neurol. 2010;9:543-‐54.
[39] Patel HC, Menon DK, Tebbs S, Hawker R, Hutchinson PJ, Kirkpatrick PJ. Specialist neurocritical care and outcome from head injury. Intensive Care Med. 2002;28:547-‐53.
[40] Mirski MA, Chang CW, Cowan R. Impact of a neuroscience intensive care unit on neurosurgical patient outcomes and cost of care: evidence-‐based support for an intensivist-‐directed specialty ICU
[44] Mushkudiani NA, Hukkelhoven CW, Hernandez AV, Murray GD, Choi SC, Maas AI, et al. A systematic review finds methodological improvements necessary for prognostic models in determining traumatic brain injury outcomes. J Clin Epidemiol. 2008;61:331-‐43.
[45] Collaborators MCT, Perel P, Arango M, Clayton T, Edwards P, Komolafe E, et al. Predicting outcome after traumatic brain injury: practical prognostic models based on large cohort of international patients. BMJ. 2008;336:425-‐9.
[46] Steyerberg EW, Mushkudiani N, Perel P, Butcher I, Lu J, McHugh GS, et al. Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics. PLoS Med. 2008;5:e165; discussion e.
[47] Roozenbeek B, Lingsma HF, Lecky FE, Lu J, Weir J, Butcher I, et al. Prediction of outcome after moderate and severe traumatic brain injury: external validation of the International Mission on Prognosis and Analysis of Clinical Trials (IMPACT) and Corticoid Randomisation After Significant Head injury (CRASH) prognostic models. Crit Care Med. 2012;40:1609-‐17.
[48] Yeoman P, Pattani H, Silcocks P, Owen V, Fuller G. Validation of the IMPACT outcome prediction score using the Nottingham Head Injury Register dataset. J Trauma. 2011;71:387-‐92.
[49] Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent Reporting of a multivariable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med. 2015;162:55-‐63.
[50] Ogundimu EO, Altman DG, Collins GS. Adequate sample size for developing prediction models is not simply related to events per variable. J Clin Epidemiol. 2016;76:175-‐82.
[51] Steyerberg EW, Mushkudiani N, Perel P, Butcher I, Lu J, McHugh GS, et al. Predicting outcome after traumatic brain injury: development and international validation of prognostic scores based on admission characteristics. PLoS Med. 2008;5:e165.
[52] Roozenbeek B, Chiu YL, Lingsma HF, Gerber LM, Steyerberg EW, Ghajar J, et al. Predicting 14-‐day mortality after severe traumatic brain injury: application of the IMPACT models in the brain trauma foundation TBI-‐trac(R) New York State database. J Neurotrauma. 2012;29:1306-‐12.
[53] Hosmer DW, Hosmer T, Le Cessie S, Lemeshow S. A comparison of goodness-‐of-‐fit tests for the logistic regression model. Stat Med. 1997;16:965-‐80.
[54] Gill M, Windemuth R, Steele R, Green SM. A comparison of the Glasgow Coma Scale score to evaluating assumptions and adequacy, and measuring and reducing errors. Stat Med. 1996;15:361-‐
87.
[57] Savitsky B, Givon A, Rozenfeld M, Radomislensky I, Peleg K. Traumatic brain injury: It is all about definition. Brain Inj. 2016;30:1194-‐200.
[58] Majdan M, Steyerberg EW, Nieboer D, Mauritz W, Rusnak M, Lingsma HF. Glasgow coma scale motor score and pupillary reaction to predict six-‐month mortality in patients with traumatic brain injury: comparison of field and admission assessment. J Neurotrauma. 2015;32:101-‐8.
[59] van Leeuwen N, Lingsma HF, Perel P, Lecky F, Roozenbeek B, Lu J, et al. Prognostic value of major extracranial injury in traumatic brain injury: an individual patient data meta-‐analysis in 39,274 patients. Neurosurgery. 2012;70:811-‐8; discussion 8.