In this study, we document the clinical effectiveness of implementing an EGDT algorithm for the management of severe sepsis and septic shock in the ED. We found a 9% absolute and 33% relative mortality reduction, suggesting a number needed to treat (1/absolute mortality reduction) of approximately 11 persons. Furthermore, this mortality benefit was found among a group of patients with at least equal hemodynamic instability based on inspection of the systolic BP and sequential organ failure scores measured at enrollment. To our knowledge, this is the largest report evaluating the clinical effectiveness of EGDT that includes prospective identification and data collection on both before implementation (before) and after implementation (after) subjects since the sentinel report by Rivers et al.7
Previous full-length reports14-17
of implementing EGDT have found improved absolute mortality rates ranging from 9 to 28%. These studies differ from the present report in that they were retrospective17
used historical control subjects,14-15
investigated bundled therapies that included EGDT,14,17
had a primary outcome other than a hospital mortality end point,16
or were cohort studies without control (preimplementation) groups.18
The present report strengthens the conclusion of these previous reports by testing a protocol very similar to that in the original report by Rivers et al.7
We submit, however, that our slight modification—namely the use of only ED staff to execute the protocol—would afford wider generalization of the protocol.
The control group in our study had a hospital mortality rate lower (27%) than previous reports7,14-17
of EGDT, which range from 29 to 53%. Our before group patients had a lower severity of illness, evidenced by the slightly lower MEDS score, than the control groups of previous studies.14-15
We speculate that the observed higher control group mortality rates in prior reports14-15,17
were secondary to systematic biases inherent to retrospective studies. Additionally, enrollment in a randomized controlled trials may have produced higher control mortality rates due to vigorous inclusion and exclusion criteria or inherent selection during the consent process.7
Our data do not allow an inference as to whether one specific component of the protocol afforded treatment effect. Subjects in the intervention group received both a significantly larger volume of crystalloid infusion and a higher frequency of vasopressor administration. We observed no significant increase in frequency of packed RBC transfusions or dobutamine administration, consistent with previous reports.14-15,17
It might be argued that a Hawthorne effect causing heightened awareness by the clinical staff (physician and nurses) resulted in both earlier and more aggressive response to physiologic abnormalities. For example, time to initial antibiotic administration decreased in the after group (142 min vs 99 min). Additionally, use of corticosteroids increased in the postintervention group (40% vs 6%), which may have impacted the results.
This study was designed to test the clinical effectiveness of EGDT as opposed to the efficacy of EGDT. An efficacy trial determines if a treatment does more good than harm when delivered under optimum conditions and in a uniform fashion.19
The efficacy of EGDT was established by Rivers et al.7
An effectiveness trial determines if a treatment does more good than harm when delivered under realworld conditions.19
Before starting this study, we chose the definition of effectiveness to be a 33% relative mortality reduction at 1 year after implementation. We chose this definition because it is the same relative mortality reduction found by Rivers et al7
in the original EGDT trial. We used 1 year as a surrogate to a numeric sample size; thus, we do not demonstrate statistically significant differences in mortality between the before and after groups.
When comparing the results of this study to that of Rivers et al,7
several points deserve comment. First, compared to the EGDT group in the study by Rivers et al,7
our after group had higher initial CVP and Scvo2
but a lower systolic BP and initial lactate concentrations than did the Rivers EGDT group. Additionally, we found a nonsignificant increase in the mean hospital length of stay (LOS) of 1.2 days (p = 0.30) and significant increase in the mean ICU LOS of 1.8 days (p < 0.05) in the after group as compared to the before group. Rivers et al7
found a nonsignificant 0.2-day difference in hospital LOS between the control and EGDT groups and did not report mean ICU LOS. Similar to Rivers et al,7
we found no significant difference in the mean duration of mechanical ventilation between our groups (p = 0.90). These data appear to indicate that we studied a less severely ill group of subjects than did Rivers et al,7
yet the protocol was still effective as indicated by the same relative mortality reduction found in both studies. Although we found no significant difference in hospital LOS, we significantly increased ICU LOS. This increase in resources utilized in the ICU is a potentially important finding that should be examined in future investigations. Finally, although we did not study a bundled protocol that included therapies other than EGDT (such as steroids and activated protein C), our after group did receive these therapies more frequently (eg
, steroids) than did the before group. This is important inasmuch as these therapies were likely not routinely administered when the original EGDT was studied was performed prior to 2001.7
Thus, when interpreting the results of our study, it is important to consider this potentially important change in routine management over time.
This report has several limitations that warrant discussion. The first limitation is a lack of a randomized design; however, this study was intended to document the clinical effectiveness of implementing the protocol into routine clinical practice rather than replicate the original study.7
Second, as previously discussed, the sample size of the study does not permit conclusions to be drawn about the statistical differences in the mortality rate between the groups. Third, it is possible that a systematic inclusion bias influenced the results of the study (eg
, in the before arm, subjects may have been missed due to underrecognition of sepsis as the etiology of illness; in the after group, patients may have been admitted with sepsis and not treated with the resuscitation protocol). Fourth, as previously mentioned, early antibiotics or increased use of steroids may have influenced the mortality difference between the groups. Also, it is possible that therapies administered after the EGDT period (eg
, during the first 72 h of ICU care) contributed to the improved outcome. Finally we did not quantify, explore, or exclude protocol deviations because this study was designed to observe the effectiveness of the protocol in a clinical, nonresearch setting. Our intention was not to quantify the number or instances where all the goals of EGDT were achieved but rather to quantify the effect of 6 h of therapy had on patient outcome.