In our study we have explored the impact of the ED treatment variables on outcome in patients admitted with suspected sepsis. We found the time to antibiotics to be an independent predictor of mortality in those who developed RH and to have a cut-off point at 55 min from arrival. The NNT for the delivery of antibiotics within 55 min of arrival, was 4. Therefore, in cases of refractory hypotension, for every four patients, the timely delivery of antibiotics saved one life. A volume > 2000 ml of IVF was found to be an independent predictor of mortality except in those with RH. A MAP ≤ 66 mmHg after 2000 ml of fluid was also found to be an independent predictor of mortality. A final MAP of ≥ 75 mmHg was associated with a lower mortality rate compared to a final MAP of < 75 mmHg but not an independent predictor of mortality.
The median time to antibiotics was 48 min and over 98% of patients received antibiotics within four hours of registering in the ED. The time to antibiotics was identified as critical in RH. The initial vital sign that would identify those who are likely to go in to RH was an initial SBP of ≤ 107 mmHg. As this was not a rounded number an initial SBP < 100 mmHg, was studied. An initial SBP < 100 mmHg had a sensitivity of 63.3% and a specificity of 88.4% for the identification of RH. The use of an initial SBP of < 100 mmHg to deliver antibiotics within an hour may reduce the number of patients receiving early and unnecessary antibiotics. However, the reduction in sensitivity for RH may be mitigated by a safety net to deliver antibiotics within 30 mins of reaching a SBP < 100 mmHg, should the SBP drop below 100 mmHg after arrival in the ED. Furthermore, we did not find the time to antibiotics to be time critical in those with a REDS score of 5 or more, a group with a mortality rate similar to those with RH.
In a multi-centre study, Seymour et al.  studied over 49,000 ED patients who had the sepsis 3 h- treatment bundle commenced within six hours of arrival in the ED. The population studied were those with severe sepsis  and septic shock. This latter group is now known as refractory hypotension. Seymour et al.  studied only those with organ dysfunction which is a sub-group of the ED population with suspected sepsis. They found the risk adjusted mortality rate for the time to antibiotics increased by an OR 1.04 for each hour delay in the administration of antibiotics. But 45% of the population studied had septic shock compared to the 5.7% satisfying the RH criteria in our study population. The findings by Seymour et al.  may be explained by the heavy weighting towards septic shock patients in their study population. The hourly odds ratio for mortality in relation to antibiotic delivery is available for those on and not on vasopressors (Seymour et al.  Supplementary file Figure S3) but not for those with and without shock. The study population categorises 22,336 patients as having septic shock although only 16,721 received vasopressors. The OR for mortality with each hour delay in antibiotic delivery was significant only in those receiving vasopressors and not significant in the 32,610 patients who did not receive vasopressors. Alam et al.  performed a multi-centre, open label, randomised controlled trial of prehospital antibiotics in sepsis. The median time to antibiotics in the ED was 70 mins and the median time to delivering antibiotics in the pre-hospital setting was 26 mins before arrival in the ED; a median 96 min between the two arms of the study. This study found that giving antibiotics in the prehospital setting did not improve outcome. Only 3.7% of this population had septic shock, a population similar to ours. These two studies concur with our finding that the timing of antibiotic delivery is critical only in those with refractory hypotension but not in the general ED population suspected of having sepsis.
We did not find the volume of fluid to be an independent predictor of mortality in those who developed RH. In fact, in this group, those who received over 2000 ml of fluid have a lower mortality rate, although this did not reach statistical significance. A clearer view could be formed when the current on-going trials [14, 15] report their findings. However, at volumes of > 2000 ml, we found intravenous fluids to be an independent predictor of mortality in those without RH, with a number-needed to harm of 14. This finding has not been reported before and would need external validation. Lane et al.  studied the association between prehospital of intravenous fluids and in-hospital mortality in patients with sepsis. They found prehospital fluids were associated with a reduction in mortality [OR 0.73 (95% CI 0.56–0.95)] in those with hypotension but not in those without hypotension. In the latter group, there was a trend towards increased mortality with prehospital fluids but this did not reach statistical significance [OR 1.41 (95% CI 0.81–2.44)]. These findings of pre-hospital fluid administration being associated with reduced mortality in those with hypotension and a trend towards increased mortality in those without hypotension, is in keeping with our study findings. As our study is a retrospective observational study, it could be argued that there may be other confounders that have not been studied may have influenced the outcome. We have tried to mitigate this by controlling the volume of fluid against the other conditions that would demand increased fluids such at the renal function and final blood pressure. However, we acknowledge that there may be other unknown factors that may have influenced the result.
Our study identified that a MAP ≤ 66 mmHg after 2000 ml of IVF was an independent predictor of mortality despite regressing it against the REDS score which includes RH. This suggests the current definition of RH, the need for vasopressors after a fluid bolus to maintain a MAP of ≥ 65 mmHg, may not be sufficient. In addition, we found a final of MAP of ≥ 75 mmHg was associated with a lower mortality rate but it was not an independent predictor or mortality. Our finding is supported by the findings by Moman et al.  who studied patients with and without acute kidney injury (AKI) in septic shock. Those without an AKI had a significantly higher median post-resuscitation MAP of 71 mmHg compared to a median MAP of 66 mmHg in those with an AKI. Asfar et al.  studied a target MAP 65–70 and 80–85 mmHg in septic shock and found no difference in mortality. But the group who should have their MAP titrated to 65–70 mmHg were actually treated to a higher MAP. These results suggest that a target MAP of ≥ 75 mmHg may be better than a MAP of ≥ 65 mmHg. However, as the final MAP was not an independent predictor of mortality in our retrospective observational study, the association with reduced mortality may be a result of a variable that we have not studied. The question whether a target MAP of ≥ 75 mmHg is better than a target MAP of ≥ 65 mmHg could be resolved by a randomised controlled trial, thus eliminating any unknown confounders.
Our study has several limitations. Firstly, it is a single centre study. Secondly, the number of patients with RH is small. Thirdly, it is a retrospective observational studies. Such studies are not ideal for the assessment of the impact of treatments as factors that influence the outcome may not have been studied. Fourthly, the appropriateness of antibiotics was not studied but it is intuitive that this may have an impact on outcome. Finally, we did not study the other treatment such as intubation and commencement of vasopressors that may have influenced outcome.