Special Feature
Does Intensivist Management Improve Patient Outcomes in The ICU? A New Study Suggests Otherwise
By James E McFeely, MD, Medical Director Critical Care Units, Alta Bates Summit Medical Center, Berkeley, CA, is Associate Editor for Critical Care Alert.
Dr. McFeely reports no financial relationship to this field of study.
Introduction: Many Previous Studies Support Management by Intensivists
A large body of literature has developed over the last decade consistently showing improved outcomes with increased use of critical care physicians in the management of patients in the ICU. While many of these studies consist of retrospective reviews with relatively modest numbers of patients, the results have all been similar, consistently showing reductions in mortality, length of stay, resource use, and/or complication rates. The results have been so similar across disciplines that fewer such articles have appeared recently, as the questions seem to have been answered satisfactorily. However, a recent article reviewing the project IMPACT national database of ICU patients has come to a startlingly different conclusion and suggests that the underlying assumptions linking improved care and outcomes to critical care trained physicians need to be revisited.1
Articles documenting improved outcomes with change to an intensivist-led ICU date back to the late 1990s. For example, one retrospective review comparing outcomes of esophageal surgery patients showed a significant reduction in mortality (14 vs 6%, p .012) and a reduction in complications (55 vs 44%, p 0.002) after changing to intensivist management in a surgical ICU.2 In another study involving a surgical ICU, the addition of daily rounds by an intensivist physician was associated with improved outcomes for abdominal aortic aneurysm patients: mortality declined (odds ratio 3.0; 95% confidence interval 1.9-4.9), and there were improvements in cardiac arrest rates, sepsis, renal failure and re-intubation rates.3
Other studies showed similar results. Adding a neurointensivist to a neurosciences ICU was associated with reductions in mortality, ICU length of stay, and rates of discharge to skilled nursing facilities.4 Similar results were seen in a medical ICU after change to a closed model of care, including reduction in hospital and ICU length of stay, decreased days on a ventilator, and no significant change in mortality.5 In the most-often cited systematic review of research in this area, published in 2002, researchers analyzed 26 relevant articles that all pointed to reduction in mortality and length of stay when the ICU goes from an open to a closed format.6
The New Study And Its Findings
Now, Mitchell, Levy and colleagues have published their analysis of the project IMPACT national database of ICU patients, examining the association between hospital mortality and management by critical care physicians.1 The IMPACT database was studied for the years 2000 through 2004, and included data on more than 140,000 patients, admitted to 123 ICUs in 100 different US hospitals. The management characteristics of the various ICUs were divided into three categories: those in which 95% of patients were managed by critical care physicians (23 units); those in which less than 5% were managed by critical care physicians (21 units); and the majority of the units falling somewhere in between (79 units), with critical care physicians available but used at the discretion of the primary physician. Approximately 30,000 patients were excluded due to missing data for variables of interest; a further 5,000 patients were excluded who had mixed styles of management during their hospital stay.
The primary outcome variable was hospital mortality. Expected mortality was determined using the SAPS II probability of mortality, modified to improve the fit with the data set. An independent score was developed using a logistic regression model to measure propensity that a patient would be selected for critical care management, along with a logistic regression subset analysis for groups of patients where a high mortality would be expected (such as patients with respiratory failure at time of admission, patients transferred from another hospital, and patients transferred from the operating room).
Some of the study's results were predictable. For example, ICUs in which critical care physicians managed 95% or more of the patients tended to be larger and exist in larger hospitals. These hospitals were also more likely to have academic affiliation and to be training hospitals for critical care fellowships. Additionally, patients managed by critical care physicians for their entire length of ICU stay tended to have a higher severity of illness as estimated by the SAPS II score than patients who did not receive critical care physician management. Patients treated by critical care physicians tended to receive more interventions (including ICU procedures, intravenous drugs, and ventilatory support), were less likely to be postoperative patients, and were more likely to be admitted from other hospitals, than in ICUs where critical care physicians were not available.
The unexpected result of this study was that actual hospital mortality was higher than predicted by the SAPS II score where patients were managed by critical care physicians and lower than predicted in ICUs where critical care physician management was not an option. The standardized mortality ratio for patients who received critical care in ICUs that managed 95% or more patients was 1.09 (95% CI, 1.05 to 1.13), compared with a standardized mortality ratio of 0.91 (CI, 0.88 to 0.94) for patients who did not receive critical care in ICUs in which critical care physicians managed 5% or fewer patients. Similar results were obtained when the data were further divided by SAPS-determined probability of death and by the propensity score that the patient was managed by a critical care physician. Even after accounting for these two measures of severity of illness, the critical care management mortality ratio was higher than predicted in almost all the subgroups analyzed. The authors' understated conclusion was that "despite adjustment for severity of illness, we cannot demonstrate any survival benefit with management by critical care physicians."1
Why Might These New Results Be Different?
These results are surprising, and at first blush are completely contradictory to the literature that preceded this article over the last ten to 20 years. Although it is difficult for the average clinician to assess and respond to these results because of the specifics of the statistical manipulation of the data, it is nevertheless clear that the authors intended to control for variables that we would all agree are relevant, such as severity of illness, patient likelihood of being selected for critical care management, and nesting of patients in specific ICU types that result in inter- and intra-ICU variability.
The authors pose several possible explanations for their results. They note that patients in the critical care management group received more ICU procedures and interventions, which in turn might be likely to result in complications which could affect mortality adversely. Another possible explanation is that despite their valiant attempts to control for co-founding variables, residual severity of illness metrics that would affect outcomes were not included in their scoring systems.
Other variables should be considered that the authors do not mention. For example, not all intensivist-managed ICUs are alike. The IMPACT database does not distinguish ICUs that have 24/7, on-site critical care physicians from those in which the patient is nominally managed by a critical care physician who makes rounds once a day, then turns over care to a hospitalist or nursing staff. Perhaps no "dose" effect was seen because many of the critical care managed units in this study had an insufficient intensivist presence to affect the mortality outcome. It may also be that hospitals that were able to participate in project IMPACT are not representative of ICUs in general. Of the thousands of ICUs in the United States, only 123 were participating in IMPACT at the time of this study. Presumably these were institutions that had both the resources and time to support data collection and pay the fee to participate in the database.
Apart from these general considerations, the most interesting group in this study is the 23 ICUs without critical care medicine coverage. These ICUs had minimal to no critical care resources available to them and yet had better than expected mortality outcomes. This subgroup of hospitals absolutely requires further study.
Possible explanations include the likelihood that these smaller ICUs tend to care for less sick patients. They tend not to have house staff and perform fewer ICU procedures, which may result in decreased complication rates. The remainder of the critical care team, if only through sheer practical necessity, may have developed techniques and procedures to make up for the lack of critical care physician input. In addition, smaller hospitals may have protocolized many aspects of ICU patient management and have better adherence to bundles of care. They may have taken the best of the critical care literature and applied it to their patient population without requiring the direct intervention of an on-site intensivist. Certainly the "low-hanging fruit" of improved ICU mortality (for example ventilator bundle, RT-directed weaning from mechanical ventilation) is available to non-intensivist physicians as well as to intensivists.
Where Do We Go from Here?
The Levy paper should provoke a substantial reassessment within the critical care community regarding the value that we bring to our hospitals and our patients. Although it is only one data point against many concerning the possibility that critical care physicians do not improve hospital mortality, it raises many questions that can and should be systematically answered. In particular it points to ICUs that are without the support of critical care medicine physicians yet appear to be doing a very good job of caring for their patients. Indeed, the most important single outcome of this article might well be the identification of this group of units and the motivation to more fully understand how they go about their business.
References
- Levy MM, et al. Association between critical care physician management and patient mortality in the intensive care unit. Ann Intern Med. 2008 Jun 3;148(11):801-9.
- Dimick JB, et al. Intensive care unit physician staffing is associated with decreased length of stay, hospital cost, and complications after esophageal resection. Crit Care Med. 2001 Apr;29(4):753-358.
- Pronovost PJ, et al. Organizational characteristics of intensive care units related to outcomes of abdominal aortic surgery. JAMA. 1999 Apr 14;281(14):1310-1317.
- Varelas, et al. J Neurosurg. 2006 May;104(5):713-719.
- Multz AS, et al. Am J Respir Crit Care Med. 1998 May;157(5 Pt 1):1468-1473.
- Pronovost PJ, et al. Physician staffing patterns and clinical outcomes in critically ill patients: a systematic review. JAMA. 2002 Nov 6;288(17):2151-2162.
You have reached your article limit for the month. Subscribe now to access this article plus other member-only content.
- Award-winning Medical Content
- Latest Advances & Development in Medicine
- Unbiased Content