Objective: Early identification and treatment of severe sepsis can significantly reduce mortality rate. We hypothesized that a risk prediction model based on early (baseline to day 1 of study) response to standard care should be significantly related to 28-day survival.
Design: Analysis of organ dysfunction data from two placebo-controlled severe sepsis trials (PROWESS and secretory phospholipase A2 inhibitor trials).
Setting: Research laboratory.
Patients: The placebo arms of two randomized, double-blind sepsis trials were combined (n = 1036). These patients met criteria for severe sepsis and received supportive standard intensive care and fluid resuscitation.
Measurements and main results: Sequential Organ Failure Assessment (SOFA) scores were calculated daily using the most aberrant physiologic or laboratory variables. Baseline and post-baseline SOFA scores categorized as improved, unchanged, or worsened were used in regression analyses correlating organ dysfunction changes with 28-day mortality. Improvement in cardiovascular (p = .0010), renal (p < .0001), or respiratory (p = .0469) function from baseline to day 1 was significantly related to survival. Odds ratios (95% confidence intervals) associated with improved vs. worsened respiratory, cardiovascular, or renal function before start of day 1 were 0.56 (0.35-0.91), 0.33 (0.18-0.59), and 0.30 (0.17-0.52), respectively. Continued improvement in cardiovascular function before start of day 2 and start of day 3 was associated with further improvement in survival (p <. 0001), with odds ratios of 0.15 (0.06-0.39) and 0.11 (0.04-0.31) for patients who improved compared with those who worsened. No other organ system was retained in the model, and improvement beyond day 1 in any other organ function did not add to the model's predictive power.
Conclusions: These analyses suggest that outcomes for patients with severe sepsis are closely related to early (baseline to day 1 here) improvement, or lack thereof, in organ function. Also, clinical improvement on subsequent days may have little additional impact on the likelihood of survival.