Introduction

It has been estimated that failure to rescue follows 17% of major complications (e.g., pneumonia, sepsis, thromboembolic events, acute renal failure, GI hemorrhage and cardiac arrest) in US hospitals [1]. ^{This translates to approximately 268,000 cases per year. Thus, the identification of critically ill inpatients and triage to an appropriate level of care is essential to provide optimal outcomes. This concept is central to the national movement towards the development of “rapid response teams,” [2] the creation of which has been met with some controversy [3]. Our goal was to evaluate the effectiveness of inpatient intensive care unit (ICU) triage as part of a needs assessment for creating a rapid response team at our institution.}

Materials and Methodology

The study was performed at the Virginia Commonwealth University Medical Center, an 820-bed, urban teaching hospital with approximately 33,000 admissions yearly. During the study period our institution did not have a rapid response or medical emergency team. At that time, when a patient’s condition worsened and the patient was deemed to possibly require ICU level care, the resident on call for the respective ICU was consulted, evaluated the patient, and then reviewed the case with a critical care or cardiology fellow. A decision regarding transfer was then made and a consultation note was placed in the medical record.

To examine the effectiveness of this ICU triage system, we evaluated ‘code blue’ events that occurred over a two-year period between 1/1/2004 and 12/31/2005 in adult patients on medical services. This was accomplished via review of our institution’s data in the National Registry of CardioPulmonary Resuscitation (NRCPR), a nationwide registry of inpatient resuscitation events sponsored by the American Heart Association. The medical records of these patients were then reviewed to determine whether a critical care consultation had occurred in the 48 hour period prior to their event. Data for these patients were abstracted to calculate SAPS II [4] ^{and Charlson scores. The SAPS II score is a severity of acute illness indicator derived from 17 variables that provides an estimate of mortality. It involves a 163-point scale, with higher scores associated with higher mortality [4, 5]. The Charlson score involves a 43-point scale that provides prognostic information based on comorbid conditions, with higher scores being associated with a higher risk of death secondary to a comorbid cause [6]. Data on code types and ultimate outcomes were collected, as well.}

Information on the total number of consults for medical ICUs for the period of interest was not available; however, we were able to obtain data on the total number of admissions to the ICUs. In order to estimate the number of consults performed, we surveyed cardiology and critical care fellows (3 of each from multiple different years of training) as to the typical number of consults and admissions that occur per day, asking them to take into account any seasonal variations that may occur. These results were then averaged and compared to the actual admissions data.

Results

A total of 367 codes occurred in our hospital from 1/1/2004 to 12/31/2005. Codes that did not occur on adult medical services were excluded, yielding a total of 89 code events. We were able to locate the medical records for 79 of these patients (88.7%). These charts were then evaluated to assess whether an ICU consultation note was placed on the chart within 48 hours prior to the ‘code blue’ event. Of the 79 patients, only 3 (3.8%) were evaluated by the ICU service and not transferred to the ICU. Table 1 summarizes the characteristics of these patients.

Figure 1

Table 1. Characteristics of code blue patients evaluated for but denied transfer to the ICU in the prior 48 hours

For the period of interest, the estimated number of medical ICU (MICU) consults was 4,080 with an estimated 2,455 admissions. The actual total number of admissions to the MICU numbered 1,409. Thus, the critical care fellows over-estimated the number of admissions by a factor of 1.74. Given this, it is not unreasonable to assume the total number of consults was similarly over-estimated. Therefore, when this is taken into account the adjusted estimated number of consults was 2,345. The estimated number of patients who received a MICU consult but were denied admission was 936 (adjusted estimated number of consults minus the actual number of admissions).

For the coronary ICU (CICU) an estimated 6,579 consults occurred with an estimated 3,655 admissions. The actual total number of admissions to the CICU numbered 2,369. The polled cardiology fellows over-estimated the number of admissions by a factor of 1.54. Again, it is not unreasonable to assume the total number of consults was similarly over-estimated. When this is taken into account the estimated number of consults was 4,272. An estimated 1,903 patients received CICU consults but were denied ICU admission.

Using the above numbers for comparison, only 2 patients (0.21%) denied admission to the MICU had a code blue event within 48 hours following the critical care consultation. Only 1 patient (0.05%) denied admission to the CICU coded within 48 hours of the critical care consultation. In toto, the rate of triage error was approximately 1 per 1,000 consultations.

Discussion

Over a two-year period, for patients on inpatient medical services at our hospital, the number of ICU consults not leading to ICU admissions was associated with a very low percentage of code blue events in the subsequent 48 hours. A review of the patients who coded after being denied ICU admission revealed that all 3 of these patients were already receiving higher than general level medical care, with 2 of the patients located in a step-down unit. The implication of this is that these patients already had been identified as being critically ill. The SAPS II and Charlson scores were highly variable, as were the ages of the patients who coded. Interestingly, all 3 of these patients suffered pulseless electrical activity arrests. All of these patients met the 2005 InterQual^® criteria [7] for admission to the respective ICUs that were consulted.

Of the 3 patients who suffered a code blue event after having been denied ICU admission, 2 (67%) survived the event itself and 1 (33%) survived to hospital discharge. For comparison, for all patients on internal medicine services who suffered a code event during the study period (89 patients), 61% survived the actual event and 22% survived until hospital discharge.

There are numerous limitations to this study. The study only involved inpatient internal medicine services in a single urban, academic hospital, and therefore the generalizability of our results beyond this setting is uncertain. Additionally, we were only able to locate the charts for approximately 89% of the patients, and therefore some cases that may have met our study criteria may have been missed. Significantly, as the study was designed to assess the effectiveness of ICU triage by looking at ‘code blue’ events, it does not comment on any morbidity or mortality associated with the timing of triage for patients who were not denied admission to ICUs, nor is it able to determine whether the patients who coded without prior consultation should have been evaluated for transfer to ICU. As well, the SAPS II score was calculated based on the 24 hour period prior to the time of the ICU consult, in an attempt to get a ‘snap-shot’ of how acutely ill these patients were; this is not the standard way to use this scoring system and therefore this affects interpretation of these values [5]. Additionally, not all of the data incorporated into this scoring system was available for all of our patients, and therefore the derived scores only reflect the information we were able to locate. Another limitation is the lack of data on number of consultations for the ICUs in question for the study period. We attempted to compensate for this by taking the information we did have, the number of admissions for the ICUs in question, and comparing this to the fellow survey data regarding total number of consultations. There is potential for both over and underestimation with this method. This likely does not significantly confound the results given the very low number of patients found to have a ‘code blue’ event after being denied ICU admission.

Conclusion

The identification and appropriate triage of critically ill patients is important in the determination of these patients’ outcomes. For inpatient medicine services in one urban, academic hospital the overall number of ICU consults not leading to ICU admissions was associated with a very low percentage of code events in the subsequent 48 hours following the time of the consult. All of these patients had been identified by their primary services as being ill enough to warrant ICU evaluation and all were hospitalized in settings receiving higher than general level medical care. These results imply that ICU consultation to evaluate for transfer to the ICU can be performed effectively with a very low triage error rate.