Lack of ICU beds has no effect on mortality ... in Canada?

What happens when the "Rapid Response Team" is called for an acutely deteriorating patient, but there's no ICU bed available to send her to? In Alberta, Canada, not much, apparently.
Henry Stelfox, Brenda Hemmelgarn, and Braden Manns analyzed 3,494 consecutive patients with "Code MET (medical emergency team)" rapid-response activations in Calgary between 2007 and 2009. They stratified them according to whether 0, 1, 2 or >2 ICU beds were available at the time, and compared their outcomes:
There was almost always an ICU bed available somewhere: Only 7% of rapid response emergency team calls occurred when no ICU beds were available. However, 14% "code METs" occurred with only 1 ICU bed available.
Patients who deteriorated without an ICU bed available were almost twice as likely to have their "goals of care" changed from resuscitation to "medical" or "comfort" care.
Mortality did not differ according to how many ICU beds were available -- it was 32-35% between the groups, p=0.82. However, the small numbers in some cohorts (e.g., only 249 METs occurred with zero ICU beds available) meant the confidence interval included a possible 25% difference in mortality between groups.
There's a seductive story here that's sure to appeal to would-be rationers, palliative care advocates, and nihilistic intensivists: maybe not transferring patients to the ICU can often result in more appropriate care. After all, if the mortality is no different between those immediately transferred and those made to wait (often never getting an ICU bed, in this analysis) after an acute deterioration on the ward, why not ration ICU care and beds? By, for example, waiting to conduct a family meeting to confirm "goals of care" before making the transfer? Although it's easy to spin this data thusly, it's not quite justified: with only 249 patients in the "acute deterioration with no ICU beds available" cohort, there's little strength in the conclusion that ICU transfer doesn't affect mortality, statistically speaking.
Too, there were almost certainly heavy cultural influences and health-system-specific factors in play here that make it even more precarious to conclude anything broadly about code METs, ICU transfer and mortality. It would be great to see a similar study done in the U.S., but it probably won't, at least not with any scale or precision: our relatively poor-quality national databases in the U.S. do not include the granular detail captured in Canada regarding code METs or many other aspects of medical care delivery.
Stelfox HT et al. Intensive Care Unit Bed Availability and Outcomes for Hospitalized Patients With Sudden Clinical Deterioration. Arch Intern Med 2012;172:467-474.Â
Scott Halpern and Jason Wagner's editorial.