Overnight intensivists unnecessary in well-staffed ICUs? (Meta-analysis)

24/7 Intensivist Coverage Does Not Improve Outcomes: Meta-Analysis

The debate over whether ICUs should be staffed around the clock by intensivist physicians has simmered for more than 20 years, with opinions on both sides driven more by values and personal biases than evidence. Recent studies have confirmed the benefit of daytime intensivists in ICUs but failed to show any additional benefit to patients if those intensivists stay overnight in the ICU. However, these studies were far from definitive, and 24/7 intensivist coverage retains intuitive, personal appeal among many.

Most ICUs in the U.S. are "open" and do not mandate intensivist consultation at all; this arrangement has been called low-intensity staffing. (Mandatory intensivist consultation, "closed" ICUs with intensivist attendings, and 24/7 intensivist coverage are considered high-intensity staffing.) Most ICUs around the world do not have 24-hour in-house physician coverage, and among those that do, the physician is usually "cross-covering" other hospital areas as well. This standard practice allows hospital systems to save money and physicians to get sleep, but during emergencies, there is no physician immediately available.

Guidelines have recommended high-intensity ICU staffing, ideally with overnight in-house coverage. Some of these were written 20 years ago, before the proliferation of multiple systems-based improvements in ICU care (i.e., protocols) and before midlevel staffing became widespread. More recently, thought leaders have renewed the call for 24/7 in-house intensivist coverage as the standard of ICU care.

Prior meta-analyses by Pronovost et al (JAMA 2002) and Young et al (Eff Clin Pract 2000) confirm the benefits of high-intensity ICU staffing over low-intensity staffing, but could not address the question of overnight intensivist coverage.

Only one randomized trial exists to guide administrators on decisions about overnight intensivist staffing: it showed no benefit from the physical presence of an intensivist overnight -- but only in a single academic ICU already covered 24/7 by residents. Multicenter observational trials are weakened by the deep inherent variation between health systems, care patterns and patient factors across geographic regions (or even across ICUs in a single hospital). The best of these observational studies, by Wallace et al (NEJM 2012), showed that high-intensity daytime staffing seemed to improve mortality, but adding in-house overnight intensivist coverage to high-intensity daytime staffing did not.

A fine new meta-analysis by M. Elizabeth Wilcox et al in the October 2013 Critical Care Medicine will give fodder to the sleep-at-home crowd. But since its substrate are the aforementioned observational trials (with all their warts), don't ask it to settle the score.

What They Did

Authors included 52 studies (all observational), enrolling 331,222 patients. Only a handful of these compared daytime high-intensity staffing against 24/7 in-house intensivist coverage. All the others compared high-intensity against low-intensity ICU staffing. Most studies used a before-and-after (historical controls) design; this is practical but particularly prone to bias (Hawthorne effect). 33 of the studies were from the U.S., the rest from around the world, and contained a diverse mix of academic, medical, surgical, and pediatric ICUs (also a few neuro and combat ICUs). Studies were of good quality and publication bias was not obviously present. Hospital mortality was only reported in 34 of the studies, and ICU mortality in only 18 studies.

What They Found

High-intensity ICU staffing was associated with 17% lower hospital mortality (RR 0.83) and 19% lower ICU mortality (RR 0.81), and a 4-hour shorter ICU stay, supporting the apparent benefit of high-intensity ICU staffing seen in previous meta-analyses.

Only six studies compared high-intensity daytime-only intensivist staffing against around-the-clock in-hospital intensivist coverage. The addition of overnight intensivists was not accompanied by an improvement in hospital mortality (RR 0.97). ICU mortality was 12% lower (RR 0.83) under 24/7 intensivist staffing, but this was not statistically significant. Numerically, this subgroup was dominated by the Wallace NEJM 2012 study.

What It Means

The question of whether overnight in-house intensivists should be the standard of care is not trivial, nor is its impact limited to lifestyle considerations. Overnight staffing would demand substantial financial investments by health systems: at a lowball estimate of one overnight physician at $400K for each of the 5,724 U.S. hospitals, that's $2.3 billion per year. Those extra intensivists don't exist today, and minting them would necessitate fundamental changes to the current training and board certification scheme, such as training hospitalists to become fast-track intensivists.

About 50% of intensivists report significant feelings of burnout (loss of enthusiasm, feeling cynical, and low sense of personal accomplishment), a number that has been consistent in surveys in France and in the U.S. The number of night shifts per month has been independently associated with burnout. On the other hand, switching to 24-hour in-hospital intensivist coverage (on a shiftwork plan) led to lower burnout scores among intensivists at 2 Canadian ICUs.

Most U.S. hospitals are community hospitals, and in most of these the ICUs are occupied (or not occupied) by patients who aren't all that sick, compared to those in ICUs in other developed countries like Canada and the U.K. (where there are 1/6th the ICU beds per capita of the U.S.). It's conceivable that overnight intensivists' contribution has a dose-response relationship, their presence helping only the sickest patients, with this signal lost in the larger pool of less-ill patients captured in the studies.

Further, in Wallace et al NEJM 2012, which dominated this meta-analysis, fully half the ICUs had 24/7 in-house resident coverage. This doesn't reflect the true mix of overnight ICU coverage in the U.S.: of 5,724 U.S. hospitals, only 400 are teaching hospitals. Many of the single-center studies included were also at academic centers (that's where the investigators work). But at non-teaching hospitals, there's no resident down the hall overnight; this is the model that needs an undiluted comparison.

Clinical Takeaway: Strong intensivist involvement saves lives among critically ill patients. If you're an attending at an academic medical center, actually being in the hospital overnight doesn't seem to help your ICU patients. In community ICUs with a strong daytime intensivist presence, things are less clear, but adding overnight coverage probably won't improve care enough to be detected in aggregated global outcomes. The decision as to whether the intuitively plausible but hypothetical benefits to individual patients justify the costs of 24/7 intensivist coverage will continue to be decided -- appropriately -- on a hospital-by-hospital basis.

Wilcox, M. Elizabeth et al. Do Intensivist Staffing Patterns Influence Hospital Mortality Following ICU Admission? A Systematic Review and Meta-Analyses. Critical Care Medicine, October 2013; 41(10): 2253-2274.