Special Feature: Avoiding Adverse Events in the ICU

By Leslie A. Hoffman, RN, PhD

Health care workers are expected to practice without error, an ideal that in reality does not occur. In fact, errors occur more frequently that we would prefer.1-3 Because the likelihood of adverse events tends to increase with the severity of illness and complexity of care, the risk for errors is especially high in critical care settings.1,3 It has been estimated that nearly all ICU patients suffer at least 1 preventable adverse event during their hospitalization.3 With publication of the Institute of Medicine Report, To Err is Human,4 which estimated that 44,000-98,000 preventable deaths occur annually in US hospitals, pressures to increase the quality of care have increased. Although these estimates have been challenged, the report highlights a problem that deserves attention. This brief essay summarizes strategies that have been advocated as a means of reducing adverse events.

Models of Error Causation

Problems due to human error can be viewed from the perspective of 2 models of error causation: the person approach and the system approach.5 Each model has a different explanation for causation and, therefore, a different approach for prevention.

The person approach focuses on unsafe acts of individuals. Adverse events are viewed as primarily arising from behaviors such as carelessness, forgetfulness, poor motivation, etc. Corrective measures are directed at changing the behavior of individuals through retraining, disciplinary actions, new or revised procedures, etc.5 There are several weaknesses of this model. First, it attributes errors to "bad behavior," rather than acknowledging that adverse events can occur even when every attempt is made to do what is correct. Second, it suggests that there is a single "root cause" for the problem. Finally, it does not promote reporting of problems due to its emphasis on personal failure.2,5

The system approach views humans as fallible and assumes that errors are to be expected, even in the best organizations. In this model, multiple defensive layers are used to minimize the likelihood of error-provoking conditions. The goal is to include sufficient barriers and safeguards to prevent adverse events from occurring. Some of these barriers are engineered, some rely on individuals, and some rely on procedures and administrative controls. A major strength of the system approach is that it encourages a "reporting culture" in which mishaps and near misses are reported so that the system can be fixed at whatever level is necessary—person, team, task, unit, institution.2,5

Most health care institutions use the person approach. Conversely, most high reliability organizations, eg, airline pilots, air traffic controllers, nuclear power plants, etc, use the system approach. Notably, these groups have few adverse events. Although the work of these groups seems remote from clinical practice, safety scientists argue that their defining cultural characteristics could be imported into health care organizations with positive benefit.5

ICU Physician Staffing

A growing body of evidence supports the assertion that quality of care in the ICU is strongly influenced by whether care is led by intensivists.6 If so, intensivist-led ICU care has the potential to decrease the number of adverse events in the unit, including those that increase mortality. Pronovost and colleagues7 reviewed findings of 26 studies conducted in North America, Europe, and Asia examining the association between ICU physician staffing and patient outcomes. ICUs were grouped into 2 categories: high-intensity (all care directed by intensivists or mandatory intensivist consultation) and low-intensity (no intensivist or elective intensivist consultation). High-intensity staffing was associated with lower hospital mortality in 16 of 17 studies (94%) (95% confidence interval [CI], 0.62-0.82) and a lower ICU mortality in 14 of 15 studies (93%) (0.61, 95% CI, 0.50-0.75). High-intensity staffing also reduced hospital stay and ICU length of stay. Study findings were consistent across a variety of populations and hospital settings. If high-intensity staffing can reduce mortality by 30%, as suggested by this review, and if it were fully implemented in all nonrural hospitals in the United States, it has been estimated that more than 160,000 lives per year could be saved.8

House Staff

Volpp and Grande9 identified 8 frequent but easily remediable problems affecting the environment in teaching institutions that predispose to adverse events. These were:

1. Frequent interruptions for paging. Frequent paging interrupts patient care, causes distractions, and is a likely source of errors. While emergencies require an immediate response, many pages are for routine matters. With alphanumeric pagers, messages could be designated as emergent, urgent, or routine. Routine communications about patients could be restricted to specific times (first 10 minutes of the hour) or sent via e-mail.9

2. Handwritten orders and elusive records. Medication errors constitute a substantial proportion of adverse events in hospitals. Illegible writing is a common culprit. Computerized medication ordering systems can prevent medication errors by scanning for potential drug interactions, a history of allergies or other contraindications, and cross-checking orders with laboratory values. More sophisticated systems can provide decision-making assistance, eg, dose alterations for age or renal disease. While some believe such systems undermine training, it can also be argued that they can have a positive training effect.9 The presence of a pharmacist on ICU rounds as a full member of the patient care team has also been shown to be effective in reducing medication errors. In one study, the rate of preventable adverse drug events decreased by 66% from 10.4 per 1000 patient days without pharmacist participation to 3.5 per 1000 patient days with the pharmacist present.10

3. Sign-out procedures. Vital information regarding a patient caseload that is transferred when another provider takes over responsibility is often conveyed in a remarkably haphazard manner, a factor that can increase errors.11 Computerized sign-out systems are available that automatically include up-to-date information on drug allergies, current medications, recent tests, relevant history, code status, and other important information. Such systems avoid problems with handwritten notes, which may be incomplete, and improve the continuity of care.9

4. Work schedules. Recent standards enacted in June 2003 mandate strict limits on the amount of time that medical trainees can work each week.12 Some have criticized this change. However, it is well established that cognitive function declines with sleep deprivation and fatigue is a common source of errors. Consequently, this mandate may reduce errors and improve, rather than decrease, efficiency.9

5. Charts and emergency equipment. Looking for charts and equipment, including emergency equipment, wastes time. Hospitals could develop standardized procedures for placement and location.

6. Academic culture. Medical training commonly uses the person approach when attempting to identify causes of adverse events, eg, the individual was not "careful enough," "didn’t try hard enough," "did not do enough reading." Consequently, house staff may be reluctant to acknowledge even the most serious errors, a behavior that prevents identifying changes that could prevent future problems.13 System flaws are best identified in an environment in which emphasis is placed on changing the system to prevent problems, rather than attributing blame.9

7. Training in procedures. The "see one, do one, teach one" approach to training is clearly not the best. Several studies have evaluated outcomes of standardized skills training with positive outcomes in academic and community settings.14,15 The wider availability of high-fidelity human simulation mannequins offers the potential of learning procedures in a safe environment and sharpening diagnostic and procedural skills in scenarios that test ability to quickly diagnose and treat a wide variety of cardiopulmonary emergencies.

8. Leadership. When care is delivered by teams of health care professionals, proficiency is enhanced by good leadership, communication, and coordination. Other industries with similar structures, eg, the aviation industry, provide formal training in team management and mandatory refresher courses. The ability to lead a team effectively can be learned by trial and error, but the time expended detracts from other responsibilities.9

ICU Nurse Staffing

A growing body of evidence suggests that adequate ICU nurse staffing is a critical determinant of patient outcomes. Understaffing has been associated with an increased risk for errors, iatrogenic complications, nosocomial infections, and death.6,16,17 In patients undergoing hepatectomy, having fewer nurses at night was associated with an increased risk for reintubation and additional costs.16 In 2606 patients undergoing abdominal aortic surgery cared for in ICUs, the likelihood of cardiac and pulmonary complications was increased by low- vs high-intensity nurse staffing.17 The increased risk of complications may reflect a difference in the level of monitoring by nurses or an insufficient number of nurses to perform interventions such as pulmonary hygiene. Given the current nursing shortage, these findings provide compelling support for efforts to improve ICU staffing.

How to Implement a System Approach

At Johns Hopkins, the use of a system approach was tested in an intervention designed to improve care of mechanically ventilated patients. In well-controlled randomized trials, several therapies have been shown to improve outcomes in mechanically ventilated patients, including elevating the head of the bed, peptic ulcer disease prophylaxis, deep vein thrombosis prevention and holding sedation so that patients can follow commands once a day. Despite evidence of benefit, Berenholtz and colleagues3 observed that use of these interventions varied widely in their ICU. To change practice, they devised a "care bundle," which included these 4 interventions. The goal was to reduce the complexity and enhance redundancy by "bundling" the 4 interventions into 1 component of care. They devised a visual tool that consisted of a ventilator with attached signs indicating the 4 measures, conducted an educational program, and developed a "daily goals" checklist to document if the therapies were provided. The checklist was completed on all patients by the ICU team during rounds, signed by the fellow or attending physician, and given to the bedside nurse before moving on to the next patient. The care bundle strategy increased the percentage of patients receiving all 4 interventions from 30% to 96%. All providers reported the checklist was easy to use, and its use has been extended to other ICUs in this health care system. This strategy is notable because it appears to have overcome barriers commonly encountered when attempting to change practice in ways that correspond with evidence-based medicine.

Facilitated incident monitoring (FIM) is a technique designed to detect adverse events that relies on voluntary, anonymous, and nonpunitive reporting of incidents by all ICU staff.18 Beckmann and colleagues18 tested the use of this approach in a 12-bed closed ICU. Medical and nursing staff were asked to identify incidents in which they were involved and to report them using an incident report form. Over the 2-month data collection period, 100 FIM reports were submitted. Nurses submitted half of the reports and half were submitted by attending physicians or housestaff. FIM identified more preventable adverse events than medical chart review and provided more useful information about the context in which the incident occurred, aspects that were helpful in identifying how to prevent future problems. FIM has the potential to facilitate system change because it promotes an organizational culture in which all team members work together to identify and eliminate potential causes of adverse events.1-3


The person model of causation has proven not effective in preventing adverse events, and there is little reason to believe that it will become more effective in the future. The system approach has several attractive characteristics. It encourages a culture in which mishaps and near misses are reported so that the system can be fixed. This approach has a long tradition of effectiveness in other high-risk organizations, such as commercial aviation and nuclear power. Preliminary evidence suggests that strategies that incorporate this approach can be successful in critical care settings. Several other strategies hold promise for reducing the number of adverse events, including staffing of ICUs by intensivists, adequate ICU nurse staffing, pharmacist participation in rounds, and changing the teaching environment in ways that reduce the risk of making errors.


1. Graf J. Crit Care Med. 2003;31:1277-1278.

2. Berwick DM. N Engl J Med. 2003;348:2570-2572.

3. Berenholtz SM, et al. Comtemp Crit Care. 2003;1(1): 1-10.

4. Institute of Medicine. To Err is Human: Building a Safer Health System. Washington, DC: National Academy Press; 1999.

5. Reason J. BMJ. 2000;320:768-770.

6. Brilli RJ, et al. Crit Care Med. 2001;29:2007-2019.

7. Pronovost PJ, et al. JAMA. 2002;288:2151-2162.

8. ICU Physician Staffing. The Leapfrog Group. Available at: www.leapfrog.org.

9. Volpp KGM, Grande D. N Engl J Med. 2003;348: 851-855.

10. Leape LL, et al. JAMA. 1999;282:267-270.

11. Petersen LA, et al. Ann Intern Med. 1994;121:866-872.

12. Accreditation Council for Graduate Medical Education (ACGME). ACGME duty hours standards now in effect for all residency programs. Available online at: http://www.acgme.org/Media/news7_1_03.asp.

13. Wu AW, et al. JAMA. 1991;265:2089-2094.

14. Sherertz RJ, et al. Ann Intern Med. 2000;132:641-648.

15. Warren DK, et al. Crit Care Med. 2003;31:1959-1963.

16. Dimick JB, et al. Am J Crit Care. 2001;10:376-382.

17. Dang D, et al. Heart Lung. 2002;31:219-228.

18. Beckman U, et al. Crit Care Med. 2003;31:1006-1011.