DOI: 10.1097/ANA.0b013e318222b5f2
,
,
PMID: 21654407
Issn Print: 0898-4921
Publication Date: 2011/07/01
Facebook
Twitter
LinkedIn
Email
 |
Checking for direct PDF access through Ovid | |
Excerpt
What are the physician's possible sources of information? Certainly, one is the physician's individual experience with similar cases1,2; another is an appraisal, as best can be done, of the physiological consequences of the lesion. Is there blood flow? Is the intracranial pressure excessively high? What is the nature of electrical activity on electroencephalogram or evoked potentials? Such information can be helpful in surmising, in a manner individualized to the specific patient, the likelihood of recovery.
However, who does not have unexpected survivors? Having experience with unexpected survivors, especially in the context of a young patient, instills an element of doubt in any scoring system when faced with an individual request for early prognostication.
Therefore, one looks for evidence in the literature of outcome statistics in groups of patients similar to the one currently at issue. However, given that many, or perhaps most, countries allow families to direct physicians to withdraw support in patients deemed hopeless, one has to wonder just how good are the data. The concern is that a publication indicates a low survival rate for a given disease, but many of the nonsurviving patients, in fact, underwent withdrawal of life support. Thus, is the cause of death from the disease or withdrawal of care? This issue was identified by Higgins et al3 as problematic in scoring systems that attempt to predict mortality. Indeed, a leading cause of death in intracerebral hemorrhage now is just that, withdrawal of life support.4 Therefore, to really be valid, one would think that patients entering a study of a tool to predict mortality would have an entry rule that support is not withdrawn—the patients progress to brain death, systemic death, or discharge from hospital to home, rehabilitation, or skilled nursing facility. Only under such a (extremely expensive) paradigm can true survival statistics be developed. If support is withdrawn early, then there is no chance for the delayed recovery, and perhaps neural regeneration, which a sentient recovery presupposes.
Ramesh et al5 in this issue of this journal may have indeed submitted just such a report. They present a sophisticated statistical model of predicting mortality in their mixed population neurosurgical ICU. However, their unit has some major contrasts when compared with most other units. The most striking difference is that there was no effort to diagnose brain death or withdraw support (commonly referred to as providing comfort care) due to their local institutional practice, even in patients unlikely to survive (based on physician perception and experience). Their prediction of mortality may actually be more accurate biologically, as patients treated in their unit are more likely to have their natural course of disease progression unaffected by physician/family decisions regarding limitation of care.
Several factors need to be considered, however, in evaluating the generalizability of the authors' predictive model. First, there must be some concern that the natural course of disease could be affected by unmeasured personal biases of treating intensivists potentially affecting their aggressiveness or enthusiasm of treatment.
