By Vibhu Sharma, MD

Attending Physician, Division of Pulmonary and Critical Care Medicine, John H. Stroger Hospital of Cook County, Assistant Professor of Medicine, Rush University Medical Center, Chicago

Dr. Sharma reports no financial relationships relevant to this field of study.

SYNOPSIS: The authors of this study used development and validation cohorts to retrospectively identify temperature trajectories over the first 72 hours from presentation in the setting of sepsis. Patients presenting with hyperthermia that resolved quickly (within the first 24 hours) had lower mortality compared to those with slow resolution or those presenting with hypothermia.

SOURCE: Bhavani SV, Carey KA, Gilbert ER, et al. Identifying novel sepsis subphenotypes using temperature trajectories. Am J Respir Crit Care Med 2019;200:327-335.

The authors of this study attempted to identify subphenotypes of sepsis based on the presenting temperature in the emergency department (ED) and the subsequent trajectory of the temperature curve over 72 hours. Patients with sepsis were selected if a blood culture order had been placed and intravenous antibiotics administered within 24 hours of presentation to the ED, defined as the time of first vital signs. Group-based trajectory modeling was used in the development cohort to assign groups based on temperature trajectories within the first 72 hours of data. Group-based trajectory modeling allows for the assessment of individual temperature patterns over time and then assigns individuals to the trajectory group with the highest membership probability. The statistical output computes groupings based on the temperature curve over time as well as individual probabilities of belonging to a specific group. Mean, maximal, and minimal temperatures were computed as well. Temperature measurements were standardized based on mean and standard deviation measurements to enable comparisons. Logistic regression was performed, with temperature trajectory grouping being the predictor variable and mortality being the outcome variable. A fever (“hyperthermia”) was defined as a temperature of > 38° C and hypothermia was defined as a temperature below 36° C.

The authors identified four different subphenotypes in the development cohort based on body temperature trajectory: 1) hyperthermic, slow resolvers (HSR); 2) hyperthermic, fast resolvers (HFR); 3) normothermic (NT); and 4) hypothermic (H). Members assigned to the HSR group presented with a fever and had no substantial change in temperature over the first 24 hours. This group also had the highest mean, maximal, and minimal temperatures during the first 72 hours. Those assigned to the HFR group presented with hyperthermia and were more likely to have their temperatures drop close to normal within the first 24 hours. Individuals assigned to the NT group remained so during the 72 hours of observation, whereas individuals assigned to the H, or hypothermic, group presented with hypothermia and stayed hypothermic for the 72-hour observation period. The temperature curves for both the development cohort and the validation cohort were remarkably similar.

The authors discovered significant mortality differences among the four groups. Logistic regression in the validation cohort revealed higher odds of mortality among those in the HSR (odds ratio [OR], 2.15; 95% confidence interval [CI], 1.77-2.61) and H (OR, 1.68; 95% CI, 1.44-1.96) groups compared to those more likely to fall in the NT group. Membership in the HFR group was protective (OR, 0.55; 95% CI, 0.44-0.68). Fever was more common in survivors, and temperature variability was higher in non-survivors compared with survivors. Erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) levels were higher in the hyperthermic groups, as were the proportion of patients with leukocytosis, attesting to immune upregulation in the hyperthermic groups. Within the hyperthermic groups, ESR and CRP, as well as the proportion of patients with leukocytosis, were higher in the HSR group compared to the HFR group. Hypothermic patients had the lowest ESR and CRP levels and the highest lactate and creatinine of the four groups. Furthermore, the hypothermic group had the highest proportion of patients requiring vasopressors (8.4%) and were the most likely to be exposed to either prednisone or methylprednisolone.

COMMENTARY

This is an interesting study that allows for consideration of temperature trajectories over time to be used to predict septic patients who may do poorly. The strength of this study lies in the large number of patients in both the development cohort and the validation cohort, as well as a tight correlation of temperature resolution curves over 72 hours in each cohort (derivation and validation) described.

The statistical technique applied to both the development and validation cohorts assigns a probability of membership of the individuals’ temperature trends to a certain group and, therefore, is not a comparison of groups. With this in mind, this study drives home the importance of assessment of a trend and not an individual temperature. For example, transient hypothermia (temperature < 36° C) was fairly common in the development cohort; 81% of patients developed one episode during the observed 72 hours. The cohorts included all patients admitted to two different institutions. While a separate cohort of critically ill patients is not identified, the analysis controlled for severity of illness and, therefore, incorporates those admitted directly to the intensive care unit (ICU) from the ED. However, the results of this study cannot be applied to patients admitted to the ICU from the floor, for example. In an attempt to homogenize populations, the authors of the study incorporated only those patients presenting to and being treated for an infection in the ED. Within this framework, patients more likely to be in the HFR group had the least exposure to vasopressors and lower mortality compared to those more likely to be in the hypothermic trajectory group, which had the highest exposure to vasopressors and higher mortality.

Recent literature has begun to assess the utility of signs assessed on physical examination of critically ill patients. For example, the authors of one study found that a fluid resuscitation strategy that targeted normalization of capillary refill time (CRT) was non-inferior to a strategy that used a lactate-driven strategy and even may have been better among those with lower severity of illness scores.1 The Simple Intensive Care Studies-I (SICS-I) group investigators focused on the predictive value of clinical signs acquired within the first 24 hours of ICU admission with respect to outcomes.2 Prolonged CRT and low peripheral temperature may predict the development of acute kidney injury (AKI).2 A particular combination of physical findings, including low central temperature, reduced urine output, and higher respiratory rate, may predict mortality as well as the APACHE score.3

The one question that the study reviewed here does not address is how temperature trajectory correlates and/or clusters with other parameters commonly assessed in tandem, such as blood pressure, respiratory rate, oxygenation indices, and possibly urine output trends over time. Based on the results of this study, it appears reasonable to investigate more aggressively those hyperthermic patients who fail to defervesce (i.e., those who could be in the HSR group) in the first 24 hours and look for persistent sources of infection when a source is apparent (e.g., pneumonia leading to meningitis or endocarditis). Similarly, patients who present with hypothermia and continue to be hypothermic over the course of hospital admission require aggressive evaluation given the increased odds of mortality. Further studies to assess how temperature trajectory interacts with the trajectory of other vital signs may allow for further subphenotyping of patients presenting with sepsis.

REFERENCES

  1. Hernandez G, Ospina-Tascon GA, Damiani LP, et al. Effect of a resuscitation strategy targeting peripheral perfusion status vs serum lactate levels on 28-day mortality among patients with septic shock: The ANDROMEDA-SHOCK randomized clinical trial. JAMA 2019;321:654-664.
  2. Wiersema R, Koeze J, Eck RJ, et al. Clinical examination findings as predictors of acute kidney injury in critically ill patients. Acta Anaesthesiol Scand 2019. doi: 10.1111/aas.13465. [Epub ahead of print].
  3. Hiemstra B, Eck RJ, Wiersema R, et al. Clinical examination for the prediction of mortality in the critically ill: The Simple Intensive Care Studies-I. Crit Care Med 2019;47:1301-1309.