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

Variation in therapy and outcome for pediatric head trauma patients

Tilford JM, Simpson PM, Yeh TS et al.

Crit Care Med 2001; 29:1056-1061. [abstract]

Reviewed by Lisa Faberowski, MD, Duke University Medical Center

Review posted March 14, 2002

I. What is being studied?

Study objective:

To examine the severity of illness on admission to the pediatric intensive care unit in pediatric head trauma patients, the therapies used during the intensive care unit stay and patient outcomes. The second objective was to assess whether insurance status or race was associated with use of therapies in pediatric head injury.

Study design

This was a prospective evaluation of consecutive admissions with the diagnosis of head trauma from three pediatric intensive care units during an 18-month period beginning June 1996.

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? Was the sample representative of the larger population they were trying to sample from?

A representative group of patients was followed since all patients were followed until time of discharge from the intensive care unit or death. All patients followed met one of these two criteria.

The termination of follow-up was either death or discharge from the pediatric intensive care unit. In terms of predictors for mortality, follow up was adequate. However, evaluation for neurologic outcome was neither complete nor noted in the results of this investigation. Follow up did not include time of discharge to home, six months or one year after injury. Thus, the study may be limited in terms of neurologic outcome scoring since continued neurologic recovery is demonstrated up until a year after the injury.

2. Were all potential predictors included?

Potential predictors were based on previous outcome investigations. Predictors of severity included physiologic data (e.g., systolic blood pressure), demographic data (e.g., age), neurologic status (e.g., GCS) and therapeutic interventions (e.g., mechanical ventilation) during the admission period. Additional therapeutic measures used during ICU admission as predictors for outcome included paralysis and sedation, anticonvulsant use, ICP monitoring, hypothermia and vasoactive drug infusion. This study is unique in that economic status and insurance class were identified in terms of predictors not only for therapy but also for outcome. Outcomes were based on the Pediatric Overall Performance Category (POPC) scale at the time of PICU discharge and ranged from a value of 1-6. One indicating normal function and 6 indicating brain death. However, there is virtually no mention of outcomes other than mortality in the results.

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

Yes, the independent contribution of each variable in the categories of resource use and severity measures was tested. Risk factors for mortality were examined by using bivariate analyses with significant predictors as candidate variables in a logistic regression to predict expected mortality.

4. Were outcome variables clearly and objectively defined?

Mortality is clearly defined. The Pediatric overall Performance Category scale was used to determine outcome. This outcome table clearly defines outcome into six categories. However, designation into one of these categories is subjective and not based on clearly defined criteria. Indeed, the designation of mild, moderate and severe disability is relative. Interestingly, the study only truly addressed mortality.

III. What are the results?

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

A univariate analysis of risk factors identified significant predictors of mortality. The predictors were then used in a multiple logistic regression model using stepwise techniques.

Table 1: Univariate analysis of predictors for mortality

Variables	Variable present % mortality	P value
Low heart rate	21.2%	<0.001
Low bicarbonate	53.1 %	<0.001
Low potassium	26.7%	<0.001
Low systolic blood pressure	40.6%	<0.001
Apnea	26.5%	<0.001
High glucose	39.2%	<0.001
Low PaCO2	30.2%	<0.001
GCS <8	30.6%	<0.001
GCS <5	32.9%	<0.001
Pupils fixed	69.0%	<0.001
Positive computed tomography scan	13.8%	<0.001
Mechanical ventilation	17.2%	<0.001
Vasoactive infusion	44.7%	<0.001
Age < 1 yr	16.1%	<0.002
Self pay	12.3%	<0.018

Table 2: Logistic regression analysis of predictors for mortality

Variable	Regression coefficient	Adjusted odds ratio (95% confidence interval)
Low bicarbonate	1.504	4.5 (1.4-14.48)
Low potassium	1.442	4.23 (1.21-14.80)
High glucose	1.281	3.6 (1.22-10.63)
Pupils fixed	2.188	8.91 (2.73-29.09)
GCS<8	1.539	4.66 (1.39-15.59)
Vasoactive infusion	1.688	5.41 (1.98-14.8)

 Calculation for mortality risk was accomplished in two steps using their model.  In the first the logit coefficients are summed for the risk factors chosen and are added to the constant term.  Then the logit equation is converted to mortality risk using the formula [pr (mortality)= e^h/1 + e^h].  The example provided by the authors is as follows:

Patient A: high glucose + GCS<8 + vasoactive infusion

h=1.281 +1.539 + 1.688 –5.065= - 0.557;

therefore, pr (mortality)= e^-0.557/1 + e^–0.557= 36%

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

The evaluation of observed and expected mortality across six levels of mortality risk indicated excellent fit (Hosmer-Lemeshow chi-square = 5.14; p = 0.273). A high level of discrimination is also noted with this model (area under the receiver operating characteristic curve = 0.94). The low death rate for patients with low probability complicated the evaluation of model fit.

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

Based on the data you can be comfortable in the mortality risk predictors as denoted by the p values < 0.001 and 95% confidence intervals and also to their concordance with previous investigations as demonstrated by the Hosmer-Lemeshaw chi-square of 5.14; p = 0.273. One must note that the low blood pressure is defined by age category (1). Further convincing is the evidence in table 2. Tables 4 and 5 evaluate risk according to PICU and to racial and insurance status.

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?

It cannot be determined from this study since the model was not used beyond this subset of patients. It can only be concluded that in this subset of patients the model worked. However, many of these parameters have been supported by previous investigations. The further utility of anticonvulsants in decreasing mortality needs to be further evaluated. The study also lacks the evaluation of neurologic outcome.

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

Given the socioeconomic variability throughout the United States, this study gives adequate representation of most patients admitted to a PICU for head injury.

3. Does the tool improve your clinical decisions?

The tool is helpful only in predicting mortality based on admission information and status. In terms of ongoing treatment modalities to decrease neurologic morbidity, this article did not address this. It is evident from this study that socioeconomic and insurance statuses have limited effect on mortality and it can be inferred that neither have an effect on therapeutic measures. The tool improves clinical decision making in that it supports the use of anticonvulsants. This is new information. The utility of this information is limited since more recent trends in therapy have used anticonvulsants as a modality for treatment not only for controlling ICP but also for seizure prophylaxis. Anticonvulsants are readily available and are not limited by cost.

Phenytoin and carbamazepine have been shown to reduce the incidence of early post traumatic seizures and thus routine use during the first seven days after head injury is currently recommended by the Brain Trauma Foundation. Valproate may have a comparable effect to phenytoin in reducing early posttraumatic seizures but is associated with a higher mortality. However, continued use of anticonvulsants beyond seven days does not prevent late posttraumatic seizures and thus continued use is not recommended (2, 3).

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

The results are not any more useful in counseling parents than information previously known. These variables essentially the data you are presented with when the child arrives to your ICU and not much can be done effect mortality. A better question to have been answered is what can we do from here to change neurologic outcome since we cannot change the variables at initial presentation.

References

1. Downard C, Hulka F, Mullins R et al. Relationship of cerebral perfusion pressure and survival in pediatric brain-injured patients. J Trauma 2000.49:654-659. [abstract]
2. Temkin NR, Dikmen SS, Wilensky AJ et al: A routine use of phenytoin for the prevention of post traumatic seizures. New Eng J Med 1990; 323:497-502. [abstract]
3. The Brain Trauma Foundation. The American Association of Neurological Surgeons. The Joint Section on Neurotrauma and Critical Care. Role of antiseizure prophylaxis following head injury. J Neurotrauma 2000;17:549-53. [abstract]

 

Comments

Back to the EB Journal Club Index