Criteria abstracted from The Users' Guide to Medical Literature, from the Health Information Research Unit and Clinical Epidemiology and Biostatistics, McMaster University

Highlighted lines and questions below provide links to the pertinent description of criteria in The EBM User's Guide, now available at the Canadian Centres for Health Evidence

Article Reviewed:

Please visit the new Evidence Based Journal Club Reviews

Survival and functional outcome in pediatric traumatic brain injury: a retrospective review and analysis of predictive factors

Hackbarth RM, Rzeszutko KM, Sturm G, et al.

Crit Care Med 2002;30(7):1630-5. [abstract; full-text for subscribers]

Reviewed by J Krishnan MD, W Morrison MD, University of Maryland, Baltimore, MD

Review posted November 24, 2002

I. What is being studied?

Study objective:

The aim of this study was to evaluate the relationship of patient care variables to survival and functional outcome in the pediatric population with traumatic brain injury(TBI)

Study design

Retrospective chart review with regression analysis to model early clinical factors predicting death and/or disability.

II. Are the results of the study valid?

Note: These questions follow from Randolph AG et al. Understanding articles describing clinical prediction tools. Crit Care Med 1998;26:1603-1612. [abstract]

1. Was a representative group of patients completely followed up? Was follow-up sufficiently long and complete?

The group of patients studied included all patients admitted to the Pediatric critical care unit(PCCU) for traumatic brain injury (TBI) over a period of four years, so it was indeed representative of the condition and outcome being studied. Since it was a chart review there was no fixed schedule for a follow up. During the study period 320 patients were identified for inclusion. Further, out of these 90 patients met the criteria for severely injured group of patients i.e. those with admission Glasgow coma scale(GCS) of < 9, those requiring ICP monitoring or mannitol, or those who died.

Fifty-two out of these 90 patients had complete data for the 52 variables and were included in the analysis. A total of ninety-five patients required inpatient rehabilitation. Twenty two of these children were < 7 and could not be categorized using the FIM score. Patients were followed until death or discharge from acute hospitalization or rehabilitation. Complete follow up was not available on all the patients due to the retrospective nature of the study. Functional outcome was measured in survivors using the FIM score; however they appropriately chose to include only children ≥ 7 in this analysis due to age related limitations in FIM scale use. The nature of the study (retrospective chart review) and the missing data limit the validity of the study.

2. Were all potential predictors included?

An effort was made to include as many predictors as possible. The authors initially collected data on 230 different variables , then narrowed it down to logical potential predictors and finally restricted the variables to those occurring in the first 24 hours after injury. The potential predictors used for analysis included variables like GCS, hypotension, bradycardia etc. in the prehospital period, hypothermia, pediatric trauma score, pH etc. in the ED, and low CPP day 1, mean CPP day 1 etc. in the Pediatric Critical Care unit.

3. Did the investigators test the independent contribution of each predictor variable?

To test for any significant relationships of potentially pertinent independent variables to survival, stepwise regression analysis was performed. First a filtering step was done to narrow the field to the most likely variables, followed by separate analyses of the remaining non PCCU and PCCU variables. A final regression analysis was carried out to analyze the remaining significant pre-hospital, emergency department (ED) and PCCU variables.

4. Were outcome variables clearly and objectively defined?

Two separate outcome variables were being studied here. The first was survival and the second was functional outcome using FIM scores. The FIM scores used were those obtained on discharge from the rehabilitation facility where 95 patients were treated. Although FIM scores may be objective criteria for the functional outcome of subject's ≥ 7yrs of age, they could not be applied to those < 7yrs of age. In addition, they do not differentiate between functional impairment secondary to TBI or an associated injury. The functional outcome data was therefore not as useful as survival.

III. What are the results?

1. What is(are) the prediction tool(s)?

During the study period 320 patients were identified for inclusion with 190 (60%) males and 130 (40%) females. 79 out of these 320 patients were classified as severely injured as per the criteria of an admitting GCS ≤ 10. Of these patients, 61 (77%) survived and 18 were non-survivors. There were no mortalities in patients with an admitting GCS of > 7. Ninety-five patients were admitted to rehabilitation; 73 were ≥ 7 years of age; of these, 70% were discharged fully independent and 27% with moderate dependence with only one patient completely dependent on others for care. Of the 21 patients ≥ 7 years of age with GCS < 7 on admission, 11 were finally discharged fully independent.

Thirteen different variables were found to be significantly associated with survival after the first step of regression analysis (table 6, page 1634) that came down to four independent variables on separating the PCCU and non-PCCU variables, namely pediatric trauma score, bradycardia and hypothermia in the ED and lowest CPP day 1 in the PCCU. These four variables were combined in a stepwise analysis that selected a model, which was explained by using two independent variables ; lowest CPP day 1 in the PCCU and bradycardia in the ED. The prediction tool derived from this exercise is an equation (page 1631) using these two variables.

2. How well does the model categorize patients into different levels of risk?

This equation gives the probability of survival. This equation has a maximal predictive value of 94% (sensitivity 98%, specificity 75%) at a probability of survival of 0.6. In other words, 3 patients of 52 (one survivor, two nonsurvivors) were incorrectly classified with this model. Although, highly significant, this model only explains 44% of the deviance. Presumably leaving more variables in the model would have strengthened its predictive ability; it is not clear why the authors chose to whittle this down to so few variables.

3. How confident are you in the estimates of risk?

The question of confidence in the estimate of risk becomes irrelevant because of the fact that this predictive value of probability of survival has never been tested in a prospective population. Their equation for predicting survival needs to be validated in a separate population (ideally prospectively) from the one in which they derived their coefficients. Clinically, their model provides enough confidence for gross prognostication, but one would never change therapy - certainly not withdraw therapy early - based on this model.

IV. Will the results help me in caring for my patients?

1. Does the tool maintain its prediction power in a new sample of patients?

This tool has never been studied in a prospective sample and hence one cannot comment on the predictive power in a different sample.

2. Are your patients similar to those patients used in deriving and validating the tool(s)?

The patient population studied is similar to that found in most academic, tertiary care PICU's.

3. Does the tool improve your clinical decisions?

The predictive tool is an equation based on two variables: cerebral perfusion pressure (CPP) and bradycardia. These two criteria, especially CPP, have already been proven in the past to be important in the management of TBI. The study confirms that maintaining CPP is very important for survival. It does not tell us the best method for maintaining adequate CPP and the authors acknowledge this. It also tells us that some children with very low GCS on admission will recover to a level of independent functioning. Therefore we should not prognosticate solely on the initial GCS.

4. Are the results useful for reassuring or counseling patients?

The predictive value of the probability calculated using this particular study may, be helpful in counseling patients provided its applicability is validated by prospective studies.

 

Comments

Back to the EB Journal Club Index