Clinical Results and Outcome Improvement Over Time in Traumatic Brain Injury

Priscilla Di Deo, Hester Lingsma, Daan Nieboer, Bob Roozenbeek, Giuseppe Citerio, Luigi Beretta, Sandra Magnoni, Elisa R. Zanier, Nino Stocchetti

Research output: Contribution to journalArticle

Abstract

Prognostic models for traumatic brain injury (TBI) are important tools both in clinical practice and research if properly validated, preferably by external validation. Prognostic models also offer the possibility of monitoring performance by comparing predicted outcomes with observed outcomes. In this study, we applied the prognostic models developed by the International Mission on Prognosis and Analysis of Clinical Trials in TBI (IMPACT) in an Italian multi-center database (Neurolink) with two aims: to compare observed with predicted outcomes and to check for a possible improvement of clinical outcome over the 11 years of patient inclusion in Neurolink. We applied the IMPACT models to patients included in Neurolink between 1997 and 2007. Performance of the models was assessed by determining calibration (with calibration plots) and discrimination (by the area under the receiver operating characteristic curve [AUC]). Logistic regression analysis was used to analyze a possible trend in outcomes over time, adjusted for predicted outcomes. A total of 1401 patients were studied. Patients had a median age of 40 years and 51% had a Glasgow Coma Scale motor score of 5 or 6. The models showed good discrimination, with AUCs of 0.86 (according to the Core Model) and 0.88 (Extended Model), and adequate calibration, with the overall observed risk of unfavorable outcome and mortality being less than predicted. Outcomes significantly improved over time. This study shows that the IMPACT models performed reasonably well in the Neurolink data and can be used for monitoring performance. After adjustment for predicted outcomes with the prognostic models, we observed a substantial improvement of patient outcomes over time in the three Neurolink centers.

Original languageEnglish
Pages (from-to)2019-2025
Number of pages7
JournalJournal of Neurotrauma
Volume33
Issue number22
DOIs
Publication statusPublished - Nov 15 2016

Fingerprint

Calibration
Clinical Trials
Area Under Curve
Glasgow Coma Scale
ROC Curve
Logistic Models
Regression Analysis
Traumatic Brain Injury
Databases
Mortality
Research

Keywords

  • outcome
  • prognostic models
  • traumatic brain injury

ASJC Scopus subject areas

  • Clinical Neurology

Cite this

Clinical Results and Outcome Improvement Over Time in Traumatic Brain Injury. / Di Deo, Priscilla; Lingsma, Hester; Nieboer, Daan; Roozenbeek, Bob; Citerio, Giuseppe; Beretta, Luigi; Magnoni, Sandra; Zanier, Elisa R.; Stocchetti, Nino.

In: Journal of Neurotrauma, Vol. 33, No. 22, 15.11.2016, p. 2019-2025.

Research output: Contribution to journalArticle

Di Deo, P, Lingsma, H, Nieboer, D, Roozenbeek, B, Citerio, G, Beretta, L, Magnoni, S, Zanier, ER & Stocchetti, N 2016, 'Clinical Results and Outcome Improvement Over Time in Traumatic Brain Injury', Journal of Neurotrauma, vol. 33, no. 22, pp. 2019-2025. https://doi.org/10.1089/neu.2015.4026
Di Deo, Priscilla ; Lingsma, Hester ; Nieboer, Daan ; Roozenbeek, Bob ; Citerio, Giuseppe ; Beretta, Luigi ; Magnoni, Sandra ; Zanier, Elisa R. ; Stocchetti, Nino. / Clinical Results and Outcome Improvement Over Time in Traumatic Brain Injury. In: Journal of Neurotrauma. 2016 ; Vol. 33, No. 22. pp. 2019-2025.
@article{d197e2d6b3f7433fad7b14461c71d4c9,
title = "Clinical Results and Outcome Improvement Over Time in Traumatic Brain Injury",
abstract = "Prognostic models for traumatic brain injury (TBI) are important tools both in clinical practice and research if properly validated, preferably by external validation. Prognostic models also offer the possibility of monitoring performance by comparing predicted outcomes with observed outcomes. In this study, we applied the prognostic models developed by the International Mission on Prognosis and Analysis of Clinical Trials in TBI (IMPACT) in an Italian multi-center database (Neurolink) with two aims: to compare observed with predicted outcomes and to check for a possible improvement of clinical outcome over the 11 years of patient inclusion in Neurolink. We applied the IMPACT models to patients included in Neurolink between 1997 and 2007. Performance of the models was assessed by determining calibration (with calibration plots) and discrimination (by the area under the receiver operating characteristic curve [AUC]). Logistic regression analysis was used to analyze a possible trend in outcomes over time, adjusted for predicted outcomes. A total of 1401 patients were studied. Patients had a median age of 40 years and 51{\%} had a Glasgow Coma Scale motor score of 5 or 6. The models showed good discrimination, with AUCs of 0.86 (according to the Core Model) and 0.88 (Extended Model), and adequate calibration, with the overall observed risk of unfavorable outcome and mortality being less than predicted. Outcomes significantly improved over time. This study shows that the IMPACT models performed reasonably well in the Neurolink data and can be used for monitoring performance. After adjustment for predicted outcomes with the prognostic models, we observed a substantial improvement of patient outcomes over time in the three Neurolink centers.",
keywords = "outcome, prognostic models, traumatic brain injury",
author = "{Di Deo}, Priscilla and Hester Lingsma and Daan Nieboer and Bob Roozenbeek and Giuseppe Citerio and Luigi Beretta and Sandra Magnoni and Zanier, {Elisa R.} and Nino Stocchetti",
year = "2016",
month = "11",
day = "15",
doi = "10.1089/neu.2015.4026",
language = "English",
volume = "33",
pages = "2019--2025",
journal = "Journal of Neurotrauma",
issn = "0897-7151",
publisher = "Mary Ann Liebert Inc.",
number = "22",

}

TY - JOUR

T1 - Clinical Results and Outcome Improvement Over Time in Traumatic Brain Injury

AU - Di Deo, Priscilla

AU - Lingsma, Hester

AU - Nieboer, Daan

AU - Roozenbeek, Bob

AU - Citerio, Giuseppe

AU - Beretta, Luigi

AU - Magnoni, Sandra

AU - Zanier, Elisa R.

AU - Stocchetti, Nino

PY - 2016/11/15

Y1 - 2016/11/15

N2 - Prognostic models for traumatic brain injury (TBI) are important tools both in clinical practice and research if properly validated, preferably by external validation. Prognostic models also offer the possibility of monitoring performance by comparing predicted outcomes with observed outcomes. In this study, we applied the prognostic models developed by the International Mission on Prognosis and Analysis of Clinical Trials in TBI (IMPACT) in an Italian multi-center database (Neurolink) with two aims: to compare observed with predicted outcomes and to check for a possible improvement of clinical outcome over the 11 years of patient inclusion in Neurolink. We applied the IMPACT models to patients included in Neurolink between 1997 and 2007. Performance of the models was assessed by determining calibration (with calibration plots) and discrimination (by the area under the receiver operating characteristic curve [AUC]). Logistic regression analysis was used to analyze a possible trend in outcomes over time, adjusted for predicted outcomes. A total of 1401 patients were studied. Patients had a median age of 40 years and 51% had a Glasgow Coma Scale motor score of 5 or 6. The models showed good discrimination, with AUCs of 0.86 (according to the Core Model) and 0.88 (Extended Model), and adequate calibration, with the overall observed risk of unfavorable outcome and mortality being less than predicted. Outcomes significantly improved over time. This study shows that the IMPACT models performed reasonably well in the Neurolink data and can be used for monitoring performance. After adjustment for predicted outcomes with the prognostic models, we observed a substantial improvement of patient outcomes over time in the three Neurolink centers.

AB - Prognostic models for traumatic brain injury (TBI) are important tools both in clinical practice and research if properly validated, preferably by external validation. Prognostic models also offer the possibility of monitoring performance by comparing predicted outcomes with observed outcomes. In this study, we applied the prognostic models developed by the International Mission on Prognosis and Analysis of Clinical Trials in TBI (IMPACT) in an Italian multi-center database (Neurolink) with two aims: to compare observed with predicted outcomes and to check for a possible improvement of clinical outcome over the 11 years of patient inclusion in Neurolink. We applied the IMPACT models to patients included in Neurolink between 1997 and 2007. Performance of the models was assessed by determining calibration (with calibration plots) and discrimination (by the area under the receiver operating characteristic curve [AUC]). Logistic regression analysis was used to analyze a possible trend in outcomes over time, adjusted for predicted outcomes. A total of 1401 patients were studied. Patients had a median age of 40 years and 51% had a Glasgow Coma Scale motor score of 5 or 6. The models showed good discrimination, with AUCs of 0.86 (according to the Core Model) and 0.88 (Extended Model), and adequate calibration, with the overall observed risk of unfavorable outcome and mortality being less than predicted. Outcomes significantly improved over time. This study shows that the IMPACT models performed reasonably well in the Neurolink data and can be used for monitoring performance. After adjustment for predicted outcomes with the prognostic models, we observed a substantial improvement of patient outcomes over time in the three Neurolink centers.

KW - outcome

KW - prognostic models

KW - traumatic brain injury

UR - http://www.scopus.com/inward/record.url?scp=84996865632&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84996865632&partnerID=8YFLogxK

U2 - 10.1089/neu.2015.4026

DO - 10.1089/neu.2015.4026

M3 - Article

AN - SCOPUS:84996865632

VL - 33

SP - 2019

EP - 2025

JO - Journal of Neurotrauma

JF - Journal of Neurotrauma

SN - 0897-7151

IS - 22

ER -