Wednesday, May 5, 2010
A couple articles about prognosis in seriously ill brain injuries in recent Neurology (see subsequent post for the second).
First is a prospective study about the accuracy of neurointensivist's prediction of neurologic prognosis for critically ill patients in a neuro ICU. It is a single institution study (Standford), in which all neuro ICU patients (being cared for by one of 5 neurointensivists) who were mechanically ventilated for over 72 hours were eligible. The neurointensivists were asked to predict the patient's 6 month functional outcome (using a modified Rankin score: 0 = no symptoms at all; 1 = no significant disability despite symptoms, able to carry out usual duties and activities; 2 = slight disability, unable to carry out all previous activities but able to look after own affairs without assistance; 3 = moderate disability, requiring some help but able to walk without assistance; 4 = moderately severe disability, unable to walk without assistance and unable to attend to own bodily needs without assist; 5 = severe disability, bedridden, incontinent, and requiring constant nursing care and attention; 6 = death. Patients or surrogates received a 6 month follow-up phone interview. For this study the authors dichotomized predicted and actual outcomes into 'good' and 'poor.' 'Poor' was a Rankin of 4-6. A prediction was considered to be correct if the qualitative range (ie 'good' vs 'poor') was correct.
The neurointensivists were were relatively young (mean years in practice after fellowship 2 years), perhaps reflecting the relatively recent rise of their practice as a discreet discipline. 144 patients were enrolled: median length of stay in ICU 12 days; median time to prediction of prognosis was 4 days from ICU admission; 50% of patients had bleeds (intracranial or subarachnoid); 18% ischemic strokes; 12% status epilepticus. 50% of the patients died by 6 months - 71% of those after a decision to discontinue life prolonging treatments. The median Rankin at 6 months (of the survivors) was 3, and 85% of survivors were available for follow-up.
Overall, the neurointensivists predicted a lower percentage of poor outcomes at 6 months than what actually occured: 54% vs 68% (with over half of those poor outcomes being death). However their predictions (good vs. poor) were accurate in 80% of cases, although more accurate when they predicted a poor outcome (94% of the cases in which they predicted poor outcomes were accurate vs 63% of the cases in which they predicted good outcomes). Excluding patients for whom life support was discontinued didn't change the accuracy of predictions by much (87% accurate for poor outcome predictions vs. 73% for good outcome predictions).
They did a few subanalyses. One of them looked at exact agreement in predicted and actual Rankin score - this exact agreement was only 43% with 74% of incorrect predictions being over-optimistic. This effect went away once patients who died were excluded. In addition to asking the docs to predict prognosis, they also asked them whether life-support should be discontinued assuming the patient's wishes were unknown. They don't disclose whether or not that was communicated to families (one assumes that within the study protocol it was not - this was just an assessment of the treating doc's opinion). They do note that 'withdrawal of care' was recommended in 27 patients and it was instituted in 23 of these - it's unclear from the paper but I think they mean that in these 27 patients the docs actually communicated a recommendation to withdraw life prolonging treatments in those 27 patients.
Besides the obvious major limitations with the paper (single institution, only 5 docs), I have a few thoughts about this. In reading the discussion at the end you get the sense that the authors aren't really sure whether to conclude that these docs were doing a good job of prognostication or not. I myself think this is basically excellent. For these patients, especially those with poor prognoses, these docs were quite accurate. ~90% is very good for any sort of 6 months prediction, and I can't think of a prognostic index for a similar patient population which is that good (but please comment if you do). At least when it comes to critically ill patients there has been some talk recently that docs tend to over-estimate how poorly patients will do (e.g. COPD patients being admitted to an ICU) and at least this study/in this population this does not seem to be the case.
One can't of course argue from this 'trust your intensivist' given the limited scope of this study; it is a reminder however of how clinicians' global assessment remains one of the best prognostic 'tools' out there (which is as much a reflection of how weak our tools are as it is how 'good' clinicians are; plus the eternal folly of thinking that it's even possible to predict the future with anymore accuracy than what they were measuring in this study).
There has also been an ongoing discussion in the neurologic literature about prognostication, and the concern that 'treatment-limiting' decisions are skewing the data. That is - in some of these studies the actual outcomes are so bad not because a certain number of the patients wouldn't have recovered but because families made decisions to limit life-prolonging treatments due to their expectation of an unacceptable outcome. These authors tried to probe that question by looking only at patients for whom no decision to withdraw life-support was made, and the accuracy of the docs' predictions remained excellent (for predictions of poor outcomes at least). That said, some of the lack of this effect is due to the dichotomous outcomes: neurointensivist predicted Rankin 4, family decide to discontinue life-prolonging treatments, patient died. For this analysis, the Rankin 4 and death are treated as the same outcome, even though the neurointensivist didn't think the patient would die from their injury....
Stepping back from those issues, one of the aspects of this sort of literature which troubles me is the clumping of a relatively heterogeneous group of patients together (into the 'poor outcome' group). This is not to argue that ending up with a Rankin score of 4 or 5 is a 'good' outcome, but that that clumping may not have the discrimination our families need in order to make decisions about what to do. There's no mention of communication ability, and in fact either Rankin 4 or 5 could describe a completely cognitively intact quadriplegic who needs however 24/7 care. Ask your local quadriplegic: many of them are perfectly fine with their 'limitations.' Given the sort of injuries described in these patients it's unlikely that really anyone ended up quadriplegic, but it's the lack of measurement of cognitive and communication ability that gives me pause about these sorts of scales. Presumably most, if not the vast majority, of patients in this study who ended up at Rankin 4, 5 had cognitive impairments, but it's just not measured or described here. While I do meet families all the time for whom physical disabilities are sufficient ('Dad said 'never let me end up in a nursing home - no matter what.'' 'If he couldn't get himself to the bathroom, he wouldn't want to go through any of this.'), just as often families 'criteria' center around cognition, and communication - the ability to recognize and interact with family, etc. And understandably so. The reverse of this of course are there are patients for whom a Rankin 3 would be unacceptable, even though it's measured here as a 'good outcome.' (In fact in one of the secondary measures of this study they asked the survivors who could communicate how they'd rate their quality of life and 38% of those with Rankin 0-3 rated it as 'fair or poor.' That's a different measure of course than whether they think it was a good thing that they lived!)
Finally, what I really wanted to know half-way through reading this, is what these docs actually said to their patients' families. Undoubtedly it was not that the patient would end up with disabilities placing them as a Rankin 5. The prognosis we formulate to ourselves may not be the one we disclose, and I've seen docs who tell me 'this patient is never going to get any better' tell families 'it's impossible to know.' I've also seen the opposite - patients who (at least in my opinion) had very indeterminate prognoses - probably bad, but I also wouldn't have been surprised if they did ok - in which language loaded with words like 'never' were used to describe their future (never walk again, never be able to feed herself).
Image from Wikipedia.
Finley Caulfield A, Gabler L, Lansberg MG, Eyngorn I, Mlynash M, Buckwalter MS, Venkatasubramanian C, & Wijman CA (2010). Outcome prediction in mechanically ventilated neurologic patients by junior neurointensivists. Neurology, 74 (14), 1096-101 PMID: 20368630